Building Internet Firewalls - Foreword
In any society, a small percentage of people are malicious. It is estimated that the Internet now has about 30 to 40 million users. Even if the percentage of malicious users is less than one percent of the overall society, the potential number of malicious users is large enough so that it should concern you.
The number of security incidents reported to the Computer Emergency Response Team Coordination Center (CERT-CC) increases every year--less than 200 in 1989, about 400 in 1991, 1400 in 1993, and 2,241 in 1994. Estimates are that we'll see more than 3000 reported incidents in 1995. Incidents occur at government and military sites, among Fortune 500 companies, at universities, and at small startups. Some incidents involve a single account on a single system. Some (for example, those involving packet sniffers) might involve as many as 100,000 systems. Of course, these numbers are only the tip of the iceberg. Many intrusions aren't reported to the CERT Coordination Center or to other computer security incident response organizations. In fact, many aren't reported at all--in some cases, because the victimized organization would rather avoid publicity or charges of carelessness, in other cases because the intrusions are not even detected.
Nobody knows the correct statistics on how many attacks are actually detected by the sites broken into, but most people in the security community agree that only a few percent are. Here's one of the few statistics I can cite: one incident response team offers a network intrusion service to its customers. With the customer's permission, they try to penetrate a system using the same tools that intruders use in their own attacks. This team found that only 4% of the sites probed detected the penetration attempts. An even more frightening estimate: Bill Cheswick of AT&T Bell Labs believes that of those attacks that do succeed, at least 40% of the attackers gain root access.[Firewalls Digest, March 31, 1995.]
It isn't only the numbers of incidents that are growing; it's the sophistication of the methods of attack. When the CERT Coordination Center was founded in the wake of the Internet worm in the fall of 1988, the attacks we faced fell into two major categories: password guessing and the exploiting of security holes in operating systems and system programs. Although too many sites still fall victim to such attacks, we're now seeing increasing technical complexity in most of the newer incidents. To some extent, this is the result of increasing consciousness among system users and system administrators--users are choosing better passwords, and administrators are applying system patches more quickly. Unfortunately, the result of this increased security consciousness isn't to stamp out security attacks; it's simply to force the attackers to learn new tricks. Many of today's attacks are more sophisticated. They include the forging of Internet Protocol (IP) addresses (intruders are guessing the sequence numbers associated with network connections and the acknowledgments between machines), the exploiting of the source routing option on IP packets on certain types of UNIX systems, and the hijacking of open terminal or login sessions.
This is not to say that all users and administrators have learned their lessons about old-style attacks. There is still more education that needs to be done among users and administrators, as well as among managers, who too often won't budget what's needed for security training. It's a dangerous world in cyberspace, and too few people realize it. This is one way in which the growth of the Internet may actually have hurt us. In the early days of the Internet, sites connected to the net usually had a whole staff of hardware and software gurus. Today, connecting to the Internet is so easy that sites forget it takes technical sophistication to connect safely and to stay secure.
Several years ago, I worked with a site in Europe that apparently had been broken into by someone who used a site in the United States to launch the attack. When I contacted the system administrator at the U.S. site, she assured me that they didn't even have computers at their location that were connected to the Internet. I told her the full domain name of the suspect system, and she replied, "Oh, you mean the Sun." It turned out that the Sun had been installed for use in a special application and had been running for years without anyone at the site realizing that it was connected to the Internet. The administrator assured me that they would disconnect the machine. Well, next morning, the European manager sent me another flaming email message--another break-in from the same U.S. system. I called the U.S. system administrator. Yes, she'd disconnected the modems. Yes, she'd disconnected the CRTs. But through it all, she'd managed to leave the system connected to the Internet. She didn't know it, but the attacker did, and he continued to take advantage of the fact.
I talk to system administrators all the time who are frustrated by break-ins, but who haven't done the basics that might prevent these break-ins from succeeding. One system administrator complained that he had reloaded his systems multiple times, and he was still being attacked. It turned out that, although he knew about CERT advisories and vendor security bulletins, he'd never bothered installing them. For example, CERT Advisory CA-93:16 was posted to the net in November of 1993; it advised the UNIX community about a problem with most versions of Sendmail. Vendors had cooperated by providing replacement programs, and the advisory contained a replacement for the /bin/sh program used in the MProg line of the sendmail.cf file. A year and a half later, CERT still gets calls from sites that are broken into using this old Sendmail vulnerability.
Although the number of security incidents continues to increase and the types of attacks become ever more sophisticated, still there is good news for those who care about security. Overall, we've seen a huge growth in awareness of the dangers of connecting to the Internet, and there's a lot of activity in the security community. One manifestation of that is the growth of the Forum of Incident Response and Security Teams (FIRST), which brings together a variety of computer security incident response teams (more than 40 at the time I'm writing this) from government, commercial, and academic organizations. Also heartening is the existence of ever-better security tools that our community makes freely available. The Computer Operations, Audit, and Security Technology (COAST) archive at Purdue is a central point for the collection and testing of many of these tools. (Appendix A of this book tells you how to contact both organizations.) Finally, the publication of some excellent books and papers on Internet security makes the hard-won wisdom of those at the front available to others.
In these dangerous times, firewalls are the best way to keep your site secure. Although you've got to include other types of security in the mix, if you're serious about connecting to the Internet, firewalls should be at the very center of your security plans. Brent Chapman has been known as the firewalls guru since the early days of firewalls on the Internet; his Firewalls mailing list and his tutorials are witness to that. Elizabeth Zwicky, especially through her work at the System Administrators Guild (SAGE) is the voice of safe and rational system administration. Together, they have written a book that will raise consciousness of, and competence in, Internet security to a new level.
CERT Technical Advisor at the CERT Coordination Center (CERT-CC)
Software Engineering Institute
Carnegie Mellon University
Great Circle Associates, Inc.
2608 Buena Vista Ave.
Alameda, CA 94501 USA
Please report problems to Webmaster@GreatCircle.COM
Copyright © 2017 Great Circle Associates, Inc.
USA Toll Free: 877 GRT CRCL
(877 478 2725)
International: +1 415 861 3588
Fax: +1 415 552 2982