Attack Detection And Secure State Estimation For Cyber-physical Systems

Cyber-physical systems(CPSs)have received widespread attention in recent years,since CPS can realize the integrated design of computing,communication,and physical systems,making the system more efficient and collaborative in real-time.As an emerging industry known as the fourth industrial revolution,CPS is a new technology revolution.It can control physical terminal equipment through information communication and calculation of wireless network,and provide a strong guarantee for the development of smart cities,intelligent transportation,and smart grids.Therefore,it has important and broad application prospects.From the structure of the cyber-physical system,it can be seen that the CPS is more and more dependent on the information network,and the openness of the network makes the attack more convenient,which brings new challenges to the security of the cyber-physical system.Consequently,attack detection and security state estimation are the first barriers to ensure the safe operation of cyber-physical systems.The cyber-physical system is a highly complex system.The information interaction between the network layer and the physical layer,or the execution of commands in the physical layer,is very timeconsuming.However,some events or commands must be responded to within a certain time.It puts forward the requirements for the real-time performance of the cyber-physical system.This property of cyber-physical systems also puts forward higher requirements for attack detection.Therefore,how to realize fast attack detection to ensure the real-time performance of cyber-physical systems has become a forward-looking direction.At the same time,due to the large-scale characteristics of CPS,how to reduce the data storage space and solve the problem of resource constraints is also a big challenge.In addition,when the system is under malicious attack,how to determine the location of the attacked sensor is also a very important direction.To sum up,this paper addresses the problem of attack detection and secure state estimation of the CPS from three aspects:ensuring the real-time of the CPS,reducing the data storage space and determining the location of the attacked node.The main contributions of this thesis are as follows:1)A novel finite time reduced order observer is designed based on the discrete cyber-physical system in which the sensors are attacked.It consists of two asymptotic observers with time delay to achieve finite time convergence.Compared with the existing finite-time observer,the finite-time reduced-order observer can effectively save the data storage space,which is very beneficial to the cyber-physical system.At the same time,combining the method of satisfiability modulus theory and the sparse observability of the system,the problem of determining the location of the attacked sensor is solved.2)Based on the continuous cyber-physical system,an impulsive reduced-order observer is proposed.Compared with the finite-time observer composed of two asymptotic observers with time delay,the proposed impulsive reduced-order observer needs less storage space to store data.At the same time,it can also well realize the finite-time observation and meet the real-time requirements of the cyber-physical system.Meanwhile,an improved satisfiability modulus theory method is proposed,which can further improve the efficiency of searching for the attacked sensor.3)Based on a cyber-physical system where sensors and actuators are simultaneously attacked,an impulsive observer with unknown input is proposed.The design of the observer solves the problem of finite-time attack detection and state estimation for the cyber-physical systems.It guarantees the real-time performance of the CPS.4)Taking a deterministic cyber-physical system—the cooperative adaptive cruise control of intelligent transportation as an example,an integral sliding mode observer is designed to detect malicious attacks when the vehicle communicates with the preceding vehicle.The safety distance is increased,and the volume ratio of the road is improved to provide a guarantee for unmanned driving.Compared with the linear sliding mode observer,the integral sliding mode observer not only maintains the robustness of the sliding mode control,but also achieves finite-time convergence.In the integral sliding mode observation design,the integral term includes a fractional power term,and its time derivative will not generate a negative fractional power,which can effectively avoid the singular problem in sliding mode control.