Optimal Byzantine Attack and Byzantine Identification in Distributed Sensor Networks
Ziteng Sun, Chuang Zhang, Pingyi Fan · 2016
Wireless sensor network is of increasing importance in data collection and event detection because of its low cost and promising performance. However, one of the drawbacks is that it is vulnerable to attacks. In this paper, we consider Byzantine attack in which byzantine sensor node deliberately commutes falsified message to degrade the system performance. We propose an optimal attacking strategy for Byzantine nodes and calculate the percentage of sensors needed to be compromised so as to blind the fusion center (FC) when there is a finite number of sensor nodes. We also develop a likelihood-based algorithm to identify Byzantine nodes using the data transmitted to the FC. The identification results are then used to further detect the presence of events. Our proposed algorithm is proved to be convergent. Simulation results show that our proposed algorithm has a better performance than reputation- based algorithms. Furthermore, the complexity of our proposed algorithm, O(T2N2), is significantly less than that of former likelihood-based algorithm, which is O(2TN2), where T is the number of events and N is the number of sensors.