Behavior-based Incentives for Node Cooperation in Wireless Ad Hoc Networks

A Mobile Ad Hoc Network (MANET) adopts a decentralized communication architecture which relies on cooperation among nodes at each layer of the protocol stack. Its reliance on cooperation for success and survival makes the ad hoc network particularly sensitive to variations in node behavior. Specifically, for functions such as routing, nodes which are limited in their resources may be unwilling to cooperate in forwarding for other nodes. Such selfish behavior leads to degradation in the performance of the network and possibly, in the extreme case, a complete cessation of operations. Consequently it is important to devise solutions to encourage resource-constrained nodes to cooperate. Incentive schemes have been proposed to induce selfish nodes to cooperate. Though many of the proposed schemes in the literature are payment-based, nodes can be incentivized to cooperate by adopting policies which are non-monetary in nature, but rather are based on the threat of retaliation for non-cooperating nodes. These policies, for which there is little formal analysis in the existing literature on node cooperation, are based on observed node behavior. We refer to them as behavior-based incentives. In this work, we analyze the effectiveness of behavior-based incentives in inducing nodes to cooperate. To determine whether an incentive scheme is effective in fostering cooperation we develop a game-theoretic model. Adopting a repeated game model, we show that nodes may agree to cooperate in sharing their resources and forward packets, even if they perceive a cost in doing so. This happens as the nodes recognize that refusing to cooperate will result in similar behavior by others, which ultimately would compromise the viability of the network as a whole. A major shortcoming in the analysis done in past works is the lack of consideration of practical constraints imposed by an ad hoc environment. One such example is the assumption that a node, when making decisions about whether to cooperate, has perfect knowledge of every other node's actions. In a distributed setting this is impractical. In our work, we analyze behavior-based incentives by incorporating such practical considerations as imperfect monitoring into our game-theoretic models. In modeling the problem as a game of imperfect public monitoring (nodes observe a common public signal that reflects the actions of other nodes in the network) we show that, under the assumption of first order stochastic dominance of the public signal, the grim trigger strategy leads to an equilibrium for nodes to cooperate. Even though a …