Sensor and Filter Attacks Against Distributed State Estimation With Dual Stealthiness

To reveal the security vulnerabilities of distributed estimation systems, we investigate the existence of false data injection attacks in this article, where attacks can corrupt sensor data and filter data. Considering the innovation-based and consensus-based strict stealthiness, a successful attack should ensure the divergence of each local estimation and not affect innovation and consensus signals. By analyzing the null space of system matrices, the necessary and sufficient conditions for the existence of attacks with such dual strict stealthiness are derived. Furthermore, if attacks can only destroy filter or sensor data, we discuss the existence of filter attacks and obtain that there is no strictly stealthy sensor attack to cause divergent estimation. By introducing a relaxed dual stealthiness constraint, the corresponding existence condition of sensor attacks is developed. Based on the given attack existence conditions, some data protection strategies are offered to defend against the threat of attacks. Simulations are conducted to demonstrate the effectiveness of the theoretical results.