Stealthy False Data Injection Attacks Against the Summation Detector in Cyber-Physical Systems

This article proposes an alternating false data injection attack strategy, which can bypass the summation detector in cyber-physical systems. This attack strategy offsets the impact on historical residuals by constantly changing the attack direction, and therefore invalidates the summation detector integrating historical information to detect the well-designed stealthy attacks. In the simulation, the proposed attack strategy reduces the increment of cumulative summation of residuals by 70% compared to the classical stealthy attack strategy, and bypasses both the $\chi ^{2}$ detector and the summation detector. Furthermore, from a more general perspective, by proposing an almost completely stealthy attack strategy to make the residual information almost unchanged, this article proves that residual based detection methods regardless of single-step residual based ones nor historical residual based ones cannot fully detect false data injection attacks.