Secure cyber-physical systems: Identification and mitigation strategies for Markovian chain-based FDI attacks

The vulnerability of cyber-physical systems (CPSs) to cyber threats such as false data injection attacks (FDIAs) bring critical challenges in terms of security and stability. afrequency deviation and area control error) which pose severe risks to the regulator mechanisms, connected equipment, and system safety. To tackle this issue, a novel defense technique with identification and mitigation is developed to eliminate the disruptive effects of FDIAs. In particular, the FDIAs are modeled by a Markovian jump process (MJP), which can threaten the system operation due to its unpredictable nature. To address this issue, the proposed defense mechanism is developed in two distinct phases, including i) an extended state observer (ESO) is utilized to recognize the occurrence of an FDIA attack with the MJP model, and ii) a mitigation-based event-triggered technique is designed to eliminate the effect of false data. In the mitigation phase, the ultra-local model control is employed to block the cyber threats using the input-output (I/O) data. To dynamically mitigate the effect of an FDIA attack, a proximal policy optimization (PPO) algorithm is utilized to adaptively adjust the gains of the established control mechanism. Extensive simulation examinations under various scenarios of FDIAs are carried out. The outcomes of the CPS system confirm the superior performance of the proposed scheme, where it outperforms prevalent techniques such as CPI controller without protection and the ANN-based Kalman filter. The improvement percentages of the proposed defense scheme (designed with the PPO tuner) for various performance indices are approximately 70 % compared to the ANN-based Kalman filter and 100 % compared to the CPI controller.