Optimal stealthy attack with historical data on cyber–physical systems

Abstract

This paper investigates the strictly stealthy attack on cyber–physical systems (CPSs), where the attacker aims to degrade the remote estimation performance maximally while bypassing the detector. A novel attack model is proposed against a general detection algorithm with arbitrary detection window length, where all the available historical data are used. Under the attack model, the achieved maximized attack performance is better than those in the existing results, and the analytical expressions of the optimal attack parameters are presented. Compared with the existing results, where the number of attack parameters is equal to the number of historical data, there is a significant parameter reduction, since only two parameters need to be designed. Finally, numerical simulations are provided to validate results.