Data Injection Attacks against Feedforward Controllers

André M. H. Teixeira

Abstract: This paper studies data injection attacks on disturbance measurements, which are forwarded to a feedforward controller. Fundamental limitations to the detectability of such attacks are examined, from which a class of stealthy attacks are characterized in terms of the zero dynamics of the plant and disturbance models. The impact of such attacks to the plant output and state estimates is also examined, in both cases of perfect and imperfect disturbance rejection by the feedforward controller. The analysis shows that, under perfect rejection, the impact of stealthy attacks on the plant output is exponentially decaying, while an exponentially increasing bias is induced on the state estimation error. The results are illustrated through numerical examples.