Data-Driven Security Control for CPSs Under Aperiodic DoS Attacks: A Switched System Approach

Abstract

This work studies a data-driven security control problem for unknown discrete-time cyber-physical systems (CPSs) under aperiodic denial-of-service (DoS) attacks based on the switched system approach. Particularly, the concerned DoS attacks are characterized by the aperiodicity with the constraints of minimum silent interval and maximum active interval. Compared with the existing data-driven methods, a novel data-driven security control approach via descriptor method and auxiliary matrices is developed for the considered CPSs with the advantage of lower computational complexity and less input-state data information. First, by utilizing the collected data and switching strategy, the considered CPSs under attacks are transformed into a direct data-driven parametrization of the corresponding switched systems. Next, a data-driven control approach via Willems’ fundamental lemma is provided. Meanwhile, through introducing auxiliary matrices to construct the connection between data information and stability conditions, a different data-driven control approach via descriptor method is also presented for the considered unknown CPSs. Finally, the efficiency and advantage of the proposed method are validated by the comparative experiment of the two-wheeled robot.