Anti-jamming Strategy Versus a Low-Power Jamming Attack When Intelligence of Adversary's Attack Type is Unknown

In this paper, we suggest a new class of anti-jamming problems where the type of intelligence associated with a jamming attack is unknown. Specifically, we consider a problem where the nodes of a peer-to-peer network do not know whether the network is under attack by a random jammer (which might be considered as a natural background noise), or an intelligent one (i.e., the jammer who can adapt his strategy based on knowledge gained during attacks). The goal of the nodes is to identify the type of the attack based on knowledge obtained from the attack in previous time slots, and thereby to reduce the efficiency of the jamming attack. First, we model the problem as a Bayesian game for a single time slot attack, and reduce it to the solution of dual linear programming (LP) problems. Additionally, the convergence of the fictitious play algorithm for finding the equilibrium is established. Then, we develop the problem for a repeated jamming attack, where the nodes adapt their beliefs based on history of the previous attacks. In particular, we have shown that it is possible for the nodes in the network to always be able to identify the jammer's type within a finite number of time slots.