Robust secure fusion estimation for cyber-physical systems under hybrid attacks

This study proposes a new theoretical framework focused on enhancing multi-sensor fusion estimation within cyber-physical systems, particularly in scenarios where sensor data may be disrupted by hybrid attacks, which involve both denial-of-service (DoS) and false data injection (FDI) attacks. The framework is built on the assumption that hybrid attacks occurrence follows a Bernoulli distribution, which informs the design of a robust secure fusion estimator aimed at minimizing estimation error variance. The framework begins by designing local robust estimators that maintain an upper bound on local estimation error covariance, even when hybrid attacks introduce parameter uncertainties. To further optimize the local estimation, parameters within these estimators are adjusted at each time step to minimize this upper bound on error covariance. To improve the accuracy of the overall system state estimation, a covariance intersection fusion strategy is applied to combine all local robust estimates into a single, more reliable fusion estimate. This fusion approach mitigates the effects of compromised sensor data and enhances resilience against hybrid attacks. The efficacy of this robust fusion estimation scheme is demonstrated through a simulation, highlighting its potential to provide secure, accurate state estimation in environments vulnerable to data manipulation.