Protocol-Based Optimal Stealthy Data-Injection Attacks via Compromised Sensors in Cyber-Physical Systems

Abstract

This article concentrates on designing a strictly stealthy attack strategy from the adversarial perspective against the multisensor cyber-physical systems (CPSs) under the round-robin protocol (RRP). The aim is to maximize the attack performance and remain strictly stealthy at the same time. Different from the existing attacks against the single-sensor application scenario, we propose a stealthy attack strategy for multisensor networked systems. In addition, due to the scheduling effects of the RRP, only incomplete transmitted data can be corrupted at each time step. Thus, a novel protocol-based attack model is developed, and then the corrupted error covariance at the remote side is derived to quantify the attack performance. Next, a convex optimization problem is formulated to solve the optimal attack parameters, which leads to the worst filtering performance. Moreover, the analytical expression of the optimal attack parameters is presented and proved. Finally, the experiments on a permanent magnet synchronous machine monitoring system are conducted to verify our results.