Event-triggered optimal stealthy false data injection attacks with side information against remote state estimation

This paper investigates optimal strictly stealthy false data injection (FDI) attacks against remote state estimation in cyber-physical systems under practical resource constraints. To efficiently allocate limited attack resources, an online innovation-dependent event-triggered attack scheduling mechanism is proposed, where attacks are launched only when the normalised intercepted innovation exceeds a predefined threshold. Once triggered, the attacker jointly exploits the intercepted innovation and side information, modelled as noisy measurements obtained from an extra sensor rather than direct access to the system state, to construct the attack signal. To evade innovation-based detectors, the tampered innovation is constrained to have the same covariance as the nominal innovation. Due to the combination of the intermittent triggering mechanism and side-information-assisted attack construction, the covariance analysis needs to simultaneously handle truncated Gaussian distributions and side-information-induced cross-covariance terms. A recursive form of the attacked estimation error covariance is analytically derived, based on which the optimal attack synthesis problem is formulated as a convex optimisation problem to maximise the trace of this covariance. Numerical examples, including an IoT-based EV simulation example derived from a physical vehicle dynamics model, demonstrate that the proposed strategy significantly degrades estimation performance while maintaining stealthiness.