Neuro-fault recognition-based cybersecurity framework for cardio-hydro-electromechanical stimulators

The modern bio-medical systems are equipped with communication networks for remote control and monitoring purposes. However, the networked systems are susceptible to cyber-attacks, which degrade system operation. This work addresses the cybersecurity problem of cardio-hydro-electromechanical stimulator (C-HEMS) under measurement tampering attacks. To do this, a safety-constrained cyber defense mechanism has been developed to ensure the security of C-HEMS system under such threats under two steps: (i) identification phase: a neuro-fault recognition scheme is proposed to identify tampering attacks by exploring the discrepancy between the observed outputs and estimated states, and (ii) alleviation phase, a non-integer model predictive control (Ni-MPC) is designed to counteract the destructive impacts of cyber anomalies in the C-HEMS. Due to the sensitive nature of the C-HEMS, a safety-enhanced deep reinforcement learning (safe-DRL) is utilized to adaptively adjust the Ni-MPC-based alleviation mechanism. By considering lower and upper bounds, the deep neural network of safe-DRL generates adaptive policies under constraint conditions. The simulation tests on the C-HEMS reveal the feasibility and accuracy of safety-resilience, based on Ni-MPC, to handle measurement tampering attacks. This bounded-policy strategy provides the possibility to dynamically avoid constraint violations, where the safeguarding of cardiac diseases will be realized by following the electrocardiogram (ECG) references under measurement tampering attacks. Quantitative results demonstrate that the proposed safety-resilient Ni-MPC improves the control performance by approximately 27.4% compared with RL-based MPC and 49.8% compared with the unprotected proportional-integral (PI) scheme, confirming its superior cardiac tracking and cyber-attack rejection capability.