A game theoretic model for stochastic routing in self-organized MANETs

In this paper, we propose a game theoretic model for zero-sum stochastic routing game in presence of the intelligent attackers in mobile ad hoc networks (MANETs). In each and every stage of the game, the source node observes the available multiple paths, the residual bandwidth of the paths, quality of the paths, and the strategy of the attackers from the previous stage of the game. Based on this observation, the source node selects an optimal routing strategy by using minimax-Q learning for secure multihop data flow from source to destination through a attack free path among the multiple paths. The selected optimal routing strategy of multihop data flow maximizes the expected payoff defined as the bandwidth utilization. The proposed technique achieves significant performance gains in terms of bandwidth utilization in MANETs. It also provides a randomized security of multihop data flow by accommodating the stochastic multihop data flow, the system dynamics, and the strategic behaviors of the attackers.