A Threshold-Based K-Means Algorithm for Selfish Behavior Detection in IEEE 802.15.4 Wireless Sensor Networks

Abstract

To overcome the insufficient of Sliding Window Non-parameter Cumulation Sum (SWN-CUSUM) algorithm, a Threshold-based K-means (TK-means) algorithm was presented to detect the selfish behaviors in an IEEE 802.15.4 Wireless Sensor Network (WSN).By judging the threshold and tracing the statistics characteristic of the average delay and the average throughput from each link, a node can determine if there is a selfish behavior and distinguish selfish nodes from normal nodes in the WSN. In the TK-means detection mechanism, when the threshold is lower than zero, it is judged that selfish behavior happens in network; In the presence of selfish behaviors, the data set is divided into K groups composed by a number of similar object, which have high similarity in the same groups and low similarity in different groups to distinguish normal nodes and selfish nodes. Simulation results from NS2 show that the TK-means algorithm can effectively detect the selfish behaviors generated by one selfish node or multiple selfish nodes in the WSN.