Gaming Terrorism
Okay, here’s the game we’re going to play: with only a limited amount of resources, you, an inspired CS2850 student, are tasked with the very important job of making sure that America is not attacked by a motley gang of well-equipped, super intelligent terrorists. The mission is to, at all times, keep watch of every airport in the small vicinity spanned by the entire fifty states. You only have a limited number of security guards at your disposal. Can you cover all the vulnerable areas with the resources you have?
For the students who have not taken this class, the immediate answer might be to shout “impossible!” After all, the situation is all but trivial to exploit if you’re playing the part of the terrorist. But for those of us who know that it is possible to use principles of game theory calculate a patrol route that gives the maximum coverage, the answer would now probably shift to a shaky “maybe.”
Fortunately for us, we don’t have to do any calculation! The scientists at the Department of Homeland Security have crunched the numbers and consumed $5 million dollars to create ARMOR, the Assistant for Randomized Monitoring Over Routes. At the heart of ARMOR is the principles of game theory: the cold hard logic behind “optimal” decision making, which in this case is the calculation of optimal patrol patterns of security guards. As the department name suggests, their goal was to figure out how to allocate our limited defensive resources to protect valuable targets.
For the algorithm to work, we are making the assumption that this “game” is described by the famous Stackelberg competition, in which each player takes turns making a move, but one player can wait until the other moves to decide what the next best move would be. By describing this situation as a “game,” the system is able to allocate defensive forces in the best possible way by looking at what can be gained vs. what can be lost. ARMOR takes into account the value of the target, the value of the target to our opponents, security tactics, time & distance, potential strategy of the attacker, economic damage, and even the loss of life, and combines it into a game in which the defensive forces make a move to best reduce negative outcome.
The fact that you can reduce a dynamic real life situation to a value-based turn-by-turn game is an impressive feat. However, for this game to work, we must also assume that the attacker is going to be acting in a completely logical way with only the goal to maximize damage. Otherwise, the machine would not have some sort of standard in which to compare its responses.
There is a very large, glaring problem with this assumption, however. The practicality of ARMOR is unknown, in real situations that aren’t restricted to the turn-by-turn rules of the game, since it is impossible to predict that the terrorists are going to make a rational move at each turn. A natural limitation exists with game theory alone, because we cannot take into account the unpredictable nature of real life.
With this is mind, the department is testing ARMOR in various other fields, such as the Coast Guard, Federal Air Marshal Service, and the TSA to make sure that the reposes provided by the system are accurate and practical. I strongly believe in this procedure, since combining the results of game theory and the tweaks provided by experimental data will ultimately lead to a more robust system.
Overall, I vote that the development of ARMOR is a good thing, because it relives the tremendous pressure of you calculating the answer to this game by hand.
http://www.defensenews.com/article/20130528/TSJ01/305280010/Game-Theory-Defeats-Terrorists