Using game theory to battle jammer in control channels of cognitive radio ad hoc networks

Because of opportunistic spectrum access, cognitive radio networks is a promising solution for spectrum scarcity problem. In cognitive radio networks, users need to communicate through control channels in order to cooperate for sharing sensing information, channel sharing and spectrum management. As a result control channels security is of a great importance. If control channels get jammed, network's performance degrades dramatically; therefore, the jammers primary target in cognitive radio networks is to demolish control channels in order to disturb the network's performance. Traditionally hopping sequences were used to battle this problem but considering cognitive radio networks intelligence, dynamic spectrum access and learning algorithms all can be hired to confront jammer. In this paper, by using game theoretic concepts, stochastic games and by defining states, actions and players' reward, we introduce a stochastic game with multiple users to make control channels resilient against sweep attacks. Win or Learn Fast (WoLF) Q learning algorithm has been used to solve the designed game and variable learning rate are used to make sure that the designed game converges. By simulation we show that in the presence of dynamic channel variations and imperfect sensing the proposed algorithm outperforms the existing ones in terms of jammer resistivity.