Dynamic Game-Theoretic Defense Approach against Stealthy Jamming Attacks in Wireless Networks

This paper develops a game-theoretic defense approach against jamming attacks targeting access points in a wireless network. We formulate a two-player zero-sum stochastic game between a network administrator (the defender) and a jammer (the attacker) in which the defender adapts the RF footprints of the nodes to counteract the jamming attack aimed at creating excessive interference in the network. Our formulation captures inherent tradeoffs between the ability of the attack to inflict damage and the attack exposure, and between reducing the interference level and maintaining network coverage. We obtain optimal policies for both players at Nash Equilibrium using a value-iteration based algorithm. To handle the state-space complexity for this class of games, we develop approximate policies by judiciously extracting features that are well-representative of the different states. Through numerical results, we show convergence of the used algorithm to stationary policies, and demonstrate the effectiveness of the defense mechanism and the approximate policies against such attacks.