Security analysis and defense strategy of distributed filtering under false data injection attacks

Abstract

This paper investigates the distributed state estimation for multi-sensor networks under false data injection attacks. The well-known χ2 detector is first considered for detecting the authenticity of the transmitted data. A necessary and sufficient condition for the insecurity of the distributed estimation system is derived under which the hostile attacks can bypass the false data detector and degrade the estimation performance. Moreover, an algorithm for generating false data is provided to keep the attack stealthy. In order to overcome the detection vulnerability, a new protection strategy is proposed to ensure that the distributed estimator is secure under false data injection attacks. It is worth emphasizing that the strategy adopts a stochastic rule instead of a fixed threshold to detect suspicious data, which effectively avoids the occurrence of the truncated Gaussian distribution. A simulation example of moving vehicle is presented to demonstrate the effectiveness of the developed approaches.