An adaptive trust system for misbehavior detection in wireless sensor networks

Abstract

Trust management has been shown to be an effective technique for protecting networks from malicious nodes and ensuring wireless sensor network (WSN) security. A number of trust systems have been proposed, but most of them are not adaptative to the current state of network security and the intensity of the attacks to which they are subjected, especially in the case of collaborative attacks. They employ fixed trust metrics derived from expert opinion rather than the objective method based on the network’s current security level. Furthermore, they are complex trust systems designed for a specific application with a high attack probability. Thus, even with a low attack rate, they consume a lot of energy. This paper proposes an adaptive trust system that considers both the network’s risk level and the trust values of sensor nodes at the same time. To match the situation in the network, the proposed system employs various trust policies. In risky situations where the WSN environment remains untrustworthy, the proposed system adjusts its trust metrics based on the network attack intensity. When the attacks are eliminated and the misbehavior rate is low, the system switches to an energy efficient policy and adjusts its trust metrics to conserve sensor node energy. Simulation results show that a zero-tolerance policy achieves 95% of the detection rate and conserves 50% of nodes’ energy under the presence of 35% of malicious nodes in the network. Energy efficient policy achieves 90% of detection rate and conserves 95% of nodes’ energy under the existence of 10% of malicious nodes in the network. Normal policy achieves up to 90% of detection rate between 15 and 25% of malicious nodes while conserving 70% and 80% of energy under these percentages.