Mata Hari and Friends (Robert Hanssen and Aldrich Ames)
Over the years, I’ve had this recurring conversation\argument with security technologists regarding the trust lifecycle. The crux of it revolves around how you go about effectively assigning, monitoring and adjusting individual trust levels. Most of us when questioned about trust will tell you that it’s made up of behavioral elements like:
- Acting with honesty and integrity
- Not having hidden agendas
- Maintaining open communication
- Keeping your promises
- Meeting your obligations
- Looking out for other people’s interests
Indeed, these are all virtuous traits, but how do we use them in designing a complex security infrastructure? After all, it’s hard to code a function that will check if a user has a hidden agenda. In order for these social concepts to be of any use, we need to understand the nature of trust; we must go "Beyond good and evil”. Under the microscope, trust exhibits the following four characteristics:
- It’s transferable—We assign a higher degree of trust to individuals who come recommended by people we already trust,
- It’s inheritable—we tend to trust a relative of a trusted friend,
- It’s socially derived—We tend to trust individuals who share our cultural heritage,
- It’s cumulative—We tend to increase our trust levels in individuals who previously have proved themselves trustworthy.
- These evaluation criteria (which, interestingly enough, are essentially deterministic Turing tests) work very well in social relationships, but frequently fail in complex security infrastructures. The source of the problem is that most of us instinctively tend to classify the world into a “friend”, “foe” or “unclassified TBD” categories. We also like to believe that once categorized, the subject in question will continue indefinitely to conform to our classification. This simplistic tendency is hard wired into our evolutionary decision making process and to a large degree also forms the basis for many irrational behaviors like anti-Semitism.
After conducting quite a few security sweeps and post mortems, I have come to conclude that most individuals—given the right opportunity and enough curiosity—could spontaneously flip the color of their “hat”.
The concept of credential-based security (that is, non-expiring clearance) is reminiscent of cheese, especially the cheap Swiss variety, the one with too many holes. Now, don’t get me wrong I have the same tolerance for curious mice as the next guy, but the text books are full of big rats that were—paradoxically—supposed to guard the cheesy comestibles, not eat or sell them! Recall that Aldrich Ames, Robert Hanssen and Kim Philby, just to name a few, each had the highest top-secret clearance and all the right personal and social attributes.
So ultimately, it’s not the rogue, external, blood thirsty anarchists or money hungry crackers one needs to worry about. Rather they are the trusted senior employees responsible for the daily maintenance, administration and security of the corporate resources. This could run the gamut from as high as the CISO who spies on the CEO’s e-mail all the way down to DBA who is running Select statements on the HR comp database.
The lesson that I have learned from all of this is that most people regardless of how trustworthy they seem, cannot be completely trusted at all times.
And you can trust me on this one.
© Copyright 2008 Yaacov Apelbaum All Rights Reserved.