Security games with interval uncertainty
Game theory has become an important tool in solving real-life decision making problems. Security games use the concept of game theory in adversarial scenarios to protect critical infrastructure. The main purpose of security games is to allocate security resources among various targets and maximize payoff for the defender considering various kinds of attackers. It is hard for domain experts to predict the attacker's behavior, so one of the major challenges in describing this game model is representing uncertainty about the attacker's payoff. Several approaches have been developed to generate these game models based on uncertainty, such as Bayesian games. However Bayesian approaches have drawbacks in solution quality and time. The work of this thesis proposes a polynomial time algorithm that represents uncertainty based on intervals and generates a robust solution for large security games unlike previous methods. I also present a methodology to transform Bayesian games with distributional uncertainty into interval games, and use this novel interval algorithm to generate an approximate solution. At the end of this thesis, empirical data shows that this novel technique is faster and generates higher quality solution compared to previous Bayesian approaches.
Artificial intelligence|Computer science
Islam, Towhidul MD., "Security games with interval uncertainty" (2013). ETD Collection for University of Texas, El Paso. AAI1545171.