A new approach to security games

Abstract

The paper proposes a new approach to finding Defender's strategy in Security Games. The method is based on a modification to the Upper Confidence bound applied to Trees (UCT) algorithm that allows to address the imperfect information games. The key advantage of our approach compared to the ones proposed hitherto in the literature lies in high flexibility of our method which can be applied, without any modifications, to a large variety of security games models. Furthermore, due to simulation-based nature of the proposed solution, various Attacker's profiles (e.g. non-completely rational behavior) can be easily tested, as opposed to the methods rooted in the game-theoretic framework.