Data‐Driven SIR‐Based Attack Detection and Defensive Control for Cyber‐Physical Systems Under Multiplicative Sensor Attacks

This paper investigates the issues of attack detection and defensive control for cyber‐physical systems (CPSs) subject to malicious sensor attacks. Unlike the common additive attacks, the overlooked multiplicative attacks are studied. Given challenges in modeling both the complex systems and the multiplicative attack dynamics, the proposed strategies are implemented within a data‐driven framework. First, the closed‐loop attacked dynamics is analyzed with the aid of coprime factorization techniques in the frequency domain. By utilizing the data‐driven stable image representations (SIRs) of the CPSs, the corrupted system output is reconstructed based on the Hankel matrices of an instrumental variable and the input‐output variables. Then, an effective tracking performance evaluator is established to detect the real‐time attack‐induced abnormal deviations. Subsequently, an active defensive feedforward controller is reconfigured to recover the degraded tracking performance using the online compromised process data. The final case study on a boost converter circuit is provided to validate the effectiveness of the proposed methods.