Security Control Of Cyber-Physical Systems Under Dynamic Data Attacks

With today's highly developed communication network and embedded technology,Cyber-Physical Systems(CPS)have entirely formed in the industry control process such as power grid and water grid,and have profound impacts on a variety of control areas.At the same time,because of the high coupling between cyber network and physical process,it will bring along with data attacks from the threat of cyber network which are willing to inject into the control process,such as virus,hacking and denial of service,and that may lead to malicious damage in physical system.Therefore,CPS urgently needs new technologies and methods to meet requirements in security control.For CPS control system with dynamic data attacks,we analyze the security control theoretically and develop corresponding algorithms to design security control scheme.Firstly,we model a CPS control system under dynamic data attacks.Considering about the possible actuator attacks and sensor attacks in control loops,we describe cyber attacks as adversarial ones in control system,and verify the vulnerable process injected by attacks with a simulation of 9-bus power grid system under attacks.Secondly,to characterize the resilience of a control system,we introduce new notions of controllability and observability to derive conditions under whether the state can be stabled and reconstructed by the existing control signal and measurements,and provide preconditions for attacks detection and resilient controller.We verify them by the proof of necessary and sufficient conditions and in a simulation and analytic methods in a simulation.Thirdly,based on the observability theorem under attacks,we optimize the state estimation method to successfully reconstructed correct states without attacks under compromised sensors,and use them as inputs for attack detection to realize the alarm.With the proof and simulation,we show the effectiveness of our proposed method.Finally,if attackers enlarge their attacked set to compromise system observability,we will focus on the actuator sides,and in the usage of controllability theorem,we develop a resilient controller to be selfadapted and switch to the optimal gain corresponding to different sets of attacks.With the proof and simulation,we show the effectiveness of our proposed method.