Improved Stealthy False Data Injection Attacks in Networked Control Systems
Chuanyi Ning, Zhiyu Xi
Abstract
This article addresses the issues of deception attacks design against state estimations and some key states in networked control systems (NCSs). Considering that the traditional stealthy attacks may lose their stealthiness in the presence of detectors with accumulative nature, the concept of improved stealthy attacks has been proposed. With the improved stealthy attack, estimation error will exhibit diverging behavior in naturally unstable systems. If the spectral radius of the system matrix is less than 1, it has been proved that the estimation errors or the divergence of key states are always bounded while suffering stealthy false data injection (FDI) attacks. Under such consideration, optimal FDI attack schemes are proposed such that the most significant impact, i.e., severest estimation error, can be achieved if stealthiness is still maintained. In addition, the problem of degrading some key states of NCSs is concerned with. Attack strategies are proposed to achieve divergence or the largest deviation of key states for systems with different spectral radii. Finally, simulation results are given to verify the feasibility and effectiveness of the proposed attack schemes.