Sliding mode-based adaptive resilient control for Markovian jump cyber–physical systems in face of simultaneous actuator and sensor attacks

Security Control of Cyber–Physical Systems under Cyber Attacks: A Survey

“…If the attack causes the actuators to perform anomalous operations, it may result in damage or the malfunctioning of the system’s physical components. This could lead to system downtime or reduced reliability, thereby impacting the normal operation of the CPS [ 31 ].…”

An adaptive control scheme for switched nonlinear cyber‐physical systems against state‐dependent sensor attacks and input‐dependent actuator attacks

“…If the parameter 𝜇 𝜎 (t) in the sensor attack model in our article satisfies 0 < 𝜇 𝜎 (t) < 1, and the parameter 𝜌 a,𝜎 (t) in the actuator attack model in our article satisfies 0 < 𝜌 a,𝜎 (t) < 1, this security control problem is similar to the sensor and actuator fault-tolerant problem. Specifically, many adaptive control schemes are proposed to reduce the impact of sensor and actuator faults in other works (see the work in References [37][38][39][40]. As for the sensor and actuator attack models considered in our article, the parameters 𝜇 𝜎 (t) and 𝜌 a,𝜎 (t) just satisfy 𝜇 𝜎 (t) ≠ −1 and 𝜌 a,𝜎 −1, which makes more drastic attacks be tolerated, and the considered models include the case of sensor and actuator faults.…”

“…Remark If the parameter $$$ {\mu}_{\sigma }(t) $$$ in the sensor attack model in our article satisfies $$$ 0<{\mu}_{\sigma }(t)<1 $$$, and the parameter $$$ {\rho}_{a,\sigma }(t) $$$ in the actuator attack model in our article satisfies $$$ 0<{\rho}_{a,\sigma }(t)<1 $$$, this security control problem is similar to the sensor and actuator fault‐tolerant problem. Specifically, many adaptive control schemes are proposed to reduce the impact of sensor and actuator faults in other works (see the work in References 37‐40). As for the sensor and actuator attack models considered in our article, the parameters $$$ {\mu}_{\sigma }(t) $$$ and $$$ {\rho}_{a,\sigma }(t) $$$ just satisfy $$$ {\mu}_{\sigma }(t)\ne -1 $$$ and $$$ {\rho}_{a,\sigma }(t)\ne -1 $$$, which makes more drastic attacks be tolerated, and the considered models include the case of sensor and actuator faults.…”

Resilient operation of DC microgrid against FDI attack: A GRU based framework

“…In order to estimate the time-series signals like voltage and current, the data-driven techniques are widely adopted because of having certain advantages [36] over classical model-based methods [8,[26][27][28][29][30][31]. Nonetheless, the classical neural network has several drawbacks, for example, it faces the random gradient explosion for deeper network while considering the long-term signal, which may lead to trapping in a locally optimal solution rather than offering a global optimum solution.…”