Bayesian Stackelberg and Security

For the life of me, I couldn't imagine why anyone would go to Iceland to participate in a Bayesian Stackelberg competition, even if it were part of the 21st European Conference on Operational Research. I mean, I've been to Iceland and I'd love to go back, but I like a little more action than seeing one Bayesian Stackelberg smash-mouth another Bayesian Stackelberg. Which shows you how much I know.

As it turns out, Bayesian Stackelberg games are serious stuff. But first, what is it? The best description I found is in Packing Trees in Communication Networks, by Mohamed Saad et al., who said that in a:

Bayeisan Stackelberg game, one player is declared as the leader and after learning her type, she first takes her action. The action taken by the leader becomes common knowledge among the followers but her type remains unknown to the followers. Followed by the action of the leader, all followers who have already learned their types take their actions simultaneously.

So why is this serious stuff? Well, Bayesian Stackelberg games are at the heart of software being used at Los Angeles International Airport to prevent terrorism. Led by Milind Tambe, a professor in artificial intelligence at the University of Southern California, researchers applied game theory to make it difficult for observers to find any patterns or regularities in vehicle security routines. The program they wrote, called ARMOR, randomizes operations without sacrificing coverage and is based on the Praveen Paruchuri's doctoral thesis entitled "Keeping the Adversary Guessing: Agent Security by Policy Randomization."

As for my trip to Iceland, it's safe to say I'll fly out of LAX.

-- Jonathan Erickson
jerickson@ddj.com

Posted by Jon Erickson at 05:48 PM  Permalink