Published in ACM SIGSAC Review, Vol. 9, No. 3, Summer 1991.
It had to be true. It was in the Washington Post. After 72 unsuccessful attempts to pass the required driving test, the District of Columbia police academy candidate was rejected in her bid to join the force. This clearly was the correct outcome. Yet I could not help but wonder, "What if she had passed on that 72nd try?" It seemed to me possible that the academy had somewhat of a lack of understanding of what it means to demonstrate basic competence.
The hope for trusted systems is what makes the National Aeronautics and Space Administration (NASA) Space Physics Applications Network (SPAN) penetrations of 1987 so dismaying. Here was a system -- Digital Equipment Corporation's (DEC's) VMS -- that had been officially evaluated and rated by the National Security Agency (NSA) and found to have an acceptable emission level. But no sooner had it been rated when some enhancements were made that had the unfortunate side effect of creating a security flaw. Some hackers learned of this when DEC issued the fix, and set about penetrating VMS systems all over the network. This turned out to be easy, since so many system managers hadn't heard of the problem or fix and hadn't installed the change when it came.4
But why dwell on this unusual case? Why not look instead at trusted systems that don't trip over their own banana peels? Actually, you'd think that effective security would be a strong selling point. That's what Gould thought, too. That's why, at the 1987 UNIX Expo show in New York, Gould advertised a free color television to anyone who could compromise the security of their UTX/32S, which also had an NSA rating. Unfortunately, the television was soon given to someone who, being clever, chose to ignore the technological defenses and distract the system manager by the technological equivalent of saying, "Hey, isn't that Dolly Parton over there?"5** Of course, from the perspective of the typical trusted system designer, this particular penetration proved nothing, because the penetrator had not played fair, but had gone around the defenses rather than through them.
Really Trusted Systems
Now by dwelling on these cases I don't want to appear too one sided. The fact is that these trusted systems I've mentioned have had only "C" ratings, which is what NSA considers to be average. Nowadays new systems are coming available that have "B" ratings, which means that their security is pretty good. In fact, these new systems are so trustworthy we plan to trust them to protect military secrets.
"What?!" you're thinking. "What?! Whose military secrets? Not ours, surely?"
But don't worry. First of all, we don't have any military secrets. But even if we did, these new systems are going to be really secure, because they're built that way from the ground up. I've talked with people working with draft versions of some of the systems. The systems aren't perfect yet, of course. There are the typical silly errors. For example, on one you could read your most highly classified mail only by logging in as uncleared. That's the old "Got it Backwards" error. On another system any user could log in at the highest security level, regardless of his clearance. But don't worry. When the final versions of these babies arrive, most of these wrinkles will have been ironed out.
The main problem with these new systems is the same problem that we have with cars: they go where you point them. This is a problem in that these new systems, like the ones that preceded them, have complex security management interfaces. The systems thus presume system managers with the demeanor of Mr. Spock (the Vulcan -- not the baby expert). At first glance, the reasons for such complex security management interfaces are difficult to fathom. What drives technologists to dream up solutions that make such faulty assumptions about human abilities, patience, diligence, and effectiveness? Are the technologists naive or just mischievous, or are they living up to their reputation as being people who can do differential calculus but can't find their car keys?
Reasons for Unreasonable Complexity
There are several reasons for these complex interfaces. The main reason is that there just is a certain irreducible amount of work and tedium involved in user registration, permission management, audit data analysis, and all those other things security managers do. Ironically, this is one area where we Americans should have an inherent advantage, in that security management is done through a TV screen. The problem with that, of course, is that the security manager channel is too boring.*** But as a result of this complexity, sometimes seemingly subtle errors in system manager etiquette can have dire consequences. This problem could be lessened by system warning labels, as are used on step ladders, but I suspect that there already is more than enough labeling going on in B systems.
Another reason for complex security management interfaces in the new, trusted systems derives from the nature of trustworthy computers. For the computer to be trustworthy, the security software has to be simple. If you add a bunch of security software to simplify the security management interface, then the system gets too complex to be trustworthy.
Now it's easy to see why readers might be a bit confused about this apparent need for trustworthy systems to have complex security management interfaces. I was confused myself 15 years ago when I first started working with the forerunner of one of these trusted systems. I even had the naivete to blurt out, "But what if the system manager just makes a mistake and labels the data wrong?" With exasperated looks, the experts slowly explained in their best parent-to-child tones:
"There's nothing we can do about that, is there? That's outside our scope. The system manager is a trusted component. If you couldn't trust the system manager, what would be the point of all this, anyway? The system manager has to be trusted."
So you can see the genesis of the idea. But since that time,
evidence has been growing to support a different idea, the idea that many
trusted humans are, well, only human, and thus imperfect. Some particularly
reckless researchers have even suggested that computers some day will
be more trustworthy and reliable than people.
The bottom line from all this is that we need to focus more attention on the system manager and on ways to simplify system management. While this might not improve the effectiveness of system managers, the attention should at least improve their sense of self.
Waiting for Poirot
Achievement of humane, secure system management clearly is a mystery that remains to be solved. Until it is, we should heed two rules:
We have to realize that, even as we improve on the security of today's systems, limited COMPUSEC effectiveness is all we can ever achieve. And maybe this is not so bad. Maybe, instead of trusting newer computers to take risks we don't dare to take now, we should instead continue to use them as we do today, and just sit back and savor the improvement. Maybe we would all benefit if, recognizing that it is not possible to build perfect defenses against each other, we concentrate instead on trying harder to get along together.
* For five years I wrote from Heidelberg, Germany, where I supported American military forces in the field. That's reality out there. And I must say that after five years of reality, it was great to move back to the Washington area.
** Actually, the technique was to write a program with a Trojan Horse in it and then ask the system manager for help with the program. When the system manager -- trying to be helpful and forgetting to be on his guard -- called up the program to look at it, the program stole his privileges.
*** Another aspect of the problem is that, even when security management is done perfectly, system security remains dependent upon the security consciousness of every single user. This is like a car in which, in addition to the main steering wheel, there also are independent steering wheels on each tire.
1. Stoll, Clifford (1989), The Cuckoo's Egg: Tracking a Spy Through the Maze of Computer Espionage, Doubleday.
2. Spafford, Eugene H. (January 1989), "The Internet Worm Program: An Analysis," ACM Computer Communication Review, Vol. 19, No. 2.
3. Alexander, Michael, Ellis Booker (26 March 1990), "Internet interloper targets hacker critics," COMPUTERWORLD.
4. Marbach, W. D., A. Nagorski, and R. Sandza (28 September 1987), "Hacking Through NASA," NEWSWEEK.
5. Smith, K. (February 1988), "Tales of the Damned," UNIX Review, Vol. 6, No. 2.
6. (Spring 1990), "An Overview," 2600 Magazine.
ACM COPYRIGHT NOTICE. Copyright © 1991 by the Association for Computing
Machinery, Inc. Permission to make digital or hard copies of part or all
of this work for personal or classroom use is granted without fee provided
that copies are not made or distributed for profit or commercial advantage
and that copies bear this notice and the full citation on the first page.
Copyrights for components of this work owned by others than ACM must be
honored. Abstracting with credit is permitted. To copy otherwise, to republish,
to post on servers, or to redistribute to lists, requires prior specific
permission and/or a fee. Request permissions from Publications Dept, ACM
Inc., fax +1 (212) 869-0481, or email@example.com.
This copy is posted by permission of ACM
and may not be redistributed.