Data-driven security controller design for unknown networked systems

This paper is concerned with security controller design of unknown networked systems under aperiodic denial-of-service (DoS) attacks, using only noise data but no model knowledge. First, a novel attack parameter-dependent stability criterion of linear networked systems under a class of time-constraint DoS attacks is proposed by using DoS attack parameter-dependent time-varying Lyapunov function method, where the considered system model, the state-feedback gain, and the lower and upper bounds of sleeping/active periods of DoS attack signal are known in advance. Based on this model-based stability condition and by combining tools from data-driven control theory, robust control theory, and switched system approach to security control, a new data-based stability criterion of all linear networked control systems (NCSs) which are consistent with the measured data and the assumed noise bound in the presence of DoS attacks is derived in terms of linear matrix inequalities. Based on this data-dependent parametrization, the data-driven security state-feedback controllers are designed correspondingly. Our control method guarantees the exponential stability properties robustly for all linear systems consistent with the measured data despite the presence of DoS attacks. As a byproduct, the proposed method embeds existing approaches for event-triggered control (ETC) into a general data-based event-triggered security control framework, which can be extended to co-design of data-based robust controller and event-triggering mechanism for uncertain NCSs under DoS attacks. Finally, the efficiency and superiority of the proposed methodology are verified through a numerical example.