Differential Kullback-Leibler Divergence Based Anomaly Detection Scheme in Sensor Networks

The constrained capacity of wireless sensor nodes and harsh, unattended deploy environments make the data collected by sensor nodes usually unreliable. In this paper, we propose a Differential Kullback-Leibler Divergence based anomaly detection scheme with the goal of detecting anomaly data values. It first partitions the whole sensor network into several clusters in which the sensors in each cluster are physically close to each other and have the similar sensed values. Then, the cluster header detects the outliers within the current cluster using Kullback-Leibler Divergence in a differential manner. The proposed scheme is lightweight and energy efficient than the existing detection scheme while maintaining the similar detection performance in terms of detection accuracy ratio and false alarm ratio.