Secure Control Of Cyber-Physical Systems Under Data Injection Attacks

With the rapid development of information science and technol-ogy and the continuous improvement of information computing and data processing capabilities,Cyber-Physical Systems(CPSs),which integrate computing units and physical objects in the network environ-ment,have gradually extended to the industry of Internet,Smart grids,intelligent transportation networks,and other control systems involving national infrastructure,industrial production and national life.High real-time,automated and intelligent support will be provided.While improving the performance and efficiency of existing systems,the open information network makes CPSs more vulnerable to the threats such as cyber attacks,malicious virus,data fraud,etc.,which leads to the vulnerability of physical system,unstable operation and even cascade accidents.Therefore,it is urgent to develop novel CPSs security defense strategies and methods to ensure healthy and stable operation.This thesis starts with the control interaction process of CPSs,and analyzes the impact of data injection attacks on system stability and observability in the network.We establish attack model,and investi-gate the stability criterion and security control method to acheive the safe operation of CPSs under attacks by combining event-triggering mechanism and state functional observer.Firstly,we analyze the general defense model of CPSs under attacks and construct the basic protection units to defend against attacks.Then a general data injection attack model existing in measurement channels and control channels is established.We also analyze the relationship between attacks and system observability.Secondly,based on the data injection attack model,we focus on the system security control problem where the attackers do not affect the system's fully observable situation.Two event-triggering mechanisms are set on the bilateral channels independently for the attack occuring on any measurement channels and control channels.The correspond-ing security control design method with known and unknown trigger thresholds is also given.The proposed event-triggering mechanisms can reduce the impact of attacks on the system by decreasing the re-sponding frequency to attacks.And the effectiveness of event-triggering dynamic output feedback control is verified by simulation.Finally,we focus on the problem where the system is partially observable under attacks.Based on reliable measurements in a certain period of time,we design a state functional observer to reconstruct the combination information of states,which can compensate for the state that cannot be observed by the attack.Using the estimated output of the functional observer,the corresponding security controller is designed,and the simulation verifies the effectiveness of the proposed method.