IEEE Access (Jan 2025)
An Observer-Based Event Triggered Mechanism for the Detection and Mitigation of FDI Attacks in Deep Brain Stimulation Systems
Abstract
Deep brain stimulation (DBS) is recognized as one of the prominent solutions to treat Parkinson’s disease. In the DBS, an accelerometer sensor is often adopted to measure the value of the tremor and then wirelessly transmit the measured signals to the programming device using a communication network. False data injection (FDI) attacks are among the most prevalent attacks in DBS, where the attacker injects false data into the communication infrastructure. In this paper, an event triggered adaptive defense mechanism is designed to mitigate the effect of FDI attacks in the DBS system. To the best of our knowledge, this is the first work focusing on the design of defense mechanisms against cybersecurity threats directed toward deep brain stimulation systems. To this end, a fractional order extended state observer (FOESO) is adopted to predict the tremor measurement signals of the accelerometer sensor. To achieve accurate prediction, Heuristic Adaptive Dynamic Programming (HADP) is utilized to adjust the coefficients embedded in the FOESO structure. Using a neural networks-based approach, the HADP agent regulates the parameters of the FOESO by maximizing a reward (reinforcement) signal according to the desired system requirements. An event-triggered mechanism is also utilized to mitigate manipulated signals measured by sensors and this strategy can address challenges such as bandwidth limitations and high network traffic. The comprehensive simulations of DBS under various levels of FDI threats reveal the feasibility of the suggested defense mechanism against cyber threats. Moreover, comparative dynamic responses and real-time examination of tremors demonstrate that by adopting the suggested adaptive defense, a superior performance than other state-of-the-art schemes can be achieved.
Keywords
