Detection of selective forwarding attacks based on adaptive learning automata and communication quality in wireless sensor networks

Abstract

Wireless sensor networks face threats of selective forwarding attacks which are simple to implement but difficult to detect. It is difficult to distinguish between malicious packet dropping and the normal packet loss on unstable wireless channels. For this situation, a selective forwarding attack detection method is proposed based on adaptive learning automata and communication quality; the method can eliminate the impact of normal packet loss on selective forwarding attack detection and can detect ordinary selective forwarding attack and special cases of selective forwarding attack. The current and comprehensive communication quality of nodes are employed to reflect the short- and long-term forwarding behaviors of nodes, and the normal packet loss caused by unstable channels and medium-access-control layer collisions is considered. The adaptive reward and penalty parameters of a detection learning automata are determined by the comprehensive communication quality of the node and the voting of its neighbors to reward normal nodes or punish malicious ones. Simulation results indicate the effectiveness of the proposed method in detecting ordinary selective forwarding attacks, black-hole attacks, on-off attacks, and energy exhaustion attacks. In addition, the communication overhead of the method is lower than that of other methods.