Data-Driven Undetectable Attack Against State Estimation in Distributed Control Systems

This article investigates a data-driven design strategy of undetectable attacks against distributed control systems. The objective of the attacker is to worsen the estimation performance and maintain undetectability through compromising partial communication links. First, the existence condition of undetectable attacks is proposed based on the null space of system matrices. Then, the subspace identification method is applied to generate the null space and construct the undetectable attack sequence utilizing the gathered system input–output data. Besides, the impact of the undetectable attack is evaluated by solving an optimization problem involving attack undetectability and energy constraints. To launch the undetectable attack at any time during the system operation, an auxiliary attack sequence is designed. Finally, the simulation results verify the effectiveness of data-driven undetectable attacks.