RSA, IBM Watson, and the Future of “Smart” Security
First of all, let me get one agenda item out of the way: What I expected to see at RSA this past week, I saw. Sadly, there wasn’t too much in the way of news – but I did see a few things that, perhaps paradoxically, relate to last week’s other high-visibility story in the IT world.
If you’ve been unconscious for the last several days, you are one of the 2 or 3 people in IT who doesn’t know that, even though Ken Jennings gave the machine a run for its money on the final day, IBM’s deep analytics and natural language processing engine known as “Watson” defeated the human champions of the game Jeopardy! Like many, I too am certain there will be application of this technology to security – even though there is clearly still a long way to go on the way toward true machine intelligence, if we ever achieve it at all.
Some have suggested that an approach such as Watson would be limited in its applicability to IT security. Computers are very good at solving finite set problems, particularly when the set is very large, but they may not be so good at things such as fraud, which poses an infinite problem set – or so some assert, at least as I understand it. I don’t agree with this entirely. For one thing, fraud isn’t so open-ended, at least in terms of its objective. Something tangible is usually at stake – like access to money, for example, or to valuable intellectual property. Even chaos can an objective, if the right targets are exploited. These can all be defined – at least in principle.
Where the problem is indeed more open-ended is that, in the real world, this definition can be difficult, to say the least. Objectives aren’t always apparent. The value of assets at risk cannot always be measured directly. Harder still to grasp are the variety of ways in which access to an objective can be obtained or exploited. Add to this the many ways in which complex IT environments constantly change – not to mention that they interact with people – and one is faced with a set of variables that are difficult to incorporate at best.
But this is exactly why I think we need to consider the value of innovation such as Watson. Though I’m not a believer in silver bullets, I feel regardless that security would benefit from Watson’s deep and fast retrieval of data coupled with an understanding of how humans seek a specific object – two fundamentals that directly address the complexity of interactions between people and technology that expose us to risk. And going a step farther: could such capability be put to work in a more dynamic approach to gamesmanship – which, after all, is one of the hallmarks of security – we may be able to deploy countermeasures that raise the challenge for the attacker above the static, predictable defenses (if any) that are the Maginot Lines of IT today (assuming that security automation itself can make similar leaps ahead).
I saw a bit of foreshadowing of this potential at RSA, among the more interesting vendor offerings at the event. Mykonos, for example, “seasons” the HTML of Web applications that end users see with feints that engage the attacker and help to expose malicious behavior. The challenges of knowing whether or not such techniques were at play could conceivably help move attacks away from such “landmined” applications and toward targets where the risk to the attacker would be less. Of course, gamesmanship would enter into this as well at some point, as attackers seek to gain an advantage over such techniques – but this is where systems that understand this sort of interaction and respond accordingly might help to keep the bar high.
Needless to say, such systems are likely still far in the future – but I sometimes wonder just how far away they really may be. Are the objectives of attackers and the potential range of tactics truly so limitless? Or are there some constraints that would make the actionable understanding of IT risk more achievable at some point? Although many disagree, rational choice theorists such as Bruce Bueno de Mesquita suggest that some seemingly random outcomes may be more predictable than many think. Can such disciplines be successfully applied to the challenges of IT security?
Perhaps…but in the mean time, we still have much to do just to get a handle on what we already know we need to do better.
- What is Watson? IBM Watson and the DeepQA Project (ibm.com)
- By Request: We Are the IBM Research Team that Developed Watson. Ask Us Anything. (reddit.com)
- How IBM’s Watson will make money (cosmiclog.msnbc.msn.com)
- IBM’s Watson Win Poses Questions on What Is Next (eweek.com)
- IBM to Design and Build Advanced Cyber Security Analytics System for the FAA (ibm.com)