Game-Theoretic Allocation of Security Investments at Nuclear Reactors.

This paper presents a game-theoretic model to guide defensive strategies for securing a nuclear facility against an attack by a highly rational and knowledgeable adversary. In our sequential play formulation, the defender chooses and implements security upgrades from a given portfolio subject to a budget constraint and in expectation of a fully informed and rational adversary. Then the adversary observes which upgrades are implemented and chooses an attack that maximizes the expected consequence. Hence, we model the situation as a two-person zero-sum game with fully symmetric information. We represent the facility as a directed graph of nodes and arcs, within which adversary capabilities are reflected via nondetection probabilities and travel times across each arc. By correlating the security force response time and probability it will defeat the adversary with the arc where the adversary is detected, we correlate the force-on-force defeat probability with the time the defender has available to prepare for the engagement. We analyze results of the model when defending against an adversary who is targeting the heat sink or offsite power supply components of a reactor. Finally, we evaluate the impact of changes in cost or effectiveness of different solutions on both the defender's optimal set of solutions and the adversary's chosen path. [ABSTRACT FROM AUTHOR]