Stealthy Attack Analysis And Design For Cyber-Physical Systems

Cyber-Physical Systems(CPSs)are a class of complex large-scale systems of integrated computing resources,communication capabilities,and physical environments that are widely used in modern industrial systems.With the development of information technology,the information space and physical space in the system are more integrated,which makes the application of information networks play a vital role in CPSs.As the advantages of the communication network are integrated into the control system,the shortcomings of the network like vulnerability and insecurity are also brought into the system.Therefore,analyzing the impact of the attacks based on the control theory plays an important role in the research on the CPSs’ security.At present,the research on the security of Cyber-physical systems is mainly from the perspective of attackers and defenders.From the perspective of defenders,researchers focus on designing corresponding detection mechanisms to offset the negative effects caused by attacks;from the perspective of attackers,the main aim is to design novel attack forms and observe the system robustness under attack.According to the difference of attack characteristics,attack forms can be divided into Stealthy Attack,Denial of Service attacks and so on.This thesis mainly considers the stealthy attack which can dodge the detection mechanism from the attacker’s point of view.Besides,this thesis analyzes the changes in system performance matrix under stealthy attacks and then gives the impact of attacks on the system state estimate.The main innovations are as follows:1)The impact of the stealthy attack on the system is analyzed,and changes in system performance matrix when there are attacks in system sensor channels and actuator channels are discussed.2)Considering a multi-sensor system with a relative entropy detector and analyzing the impact of attacks on system performance.Based on the Kullback-Leibler Divergence,the maximum attack form which can be added to the system’s innovation is analyzed and the numerical solution of the magnitude of the worst-case stealthy attack probability is given,and the trade-off relationship between the attack and the detector threshold is explained.3)The security problem of multi-sensor fusion systems equipped with decentralized Kalman filters is studied.Based on the mathematical expectation,the state estimation error covariance expression of the system under linear stealthy attack is derived.Furthermore,the quantitative relationship between the fusion system and the subsystem error covariance and also the optimal attack strategy are given.