A Study On State Estimation Based Attack And Defense Strategies For Cyber Physical Systems

As an integration of sensing,computing,communication,control and physical processing,Cyber-Physical Systems(CPSs)are widely used in industrial control,smart grid,medical equipment,etc.In practice,CPSs often suffer from malicious attacks,such as False Data Injection(FDI),interference,eavesdropping,etc.,thereby being hard to work normally.As an important basis for control decision-making,state estimation contributes to a better understanding of security of CPSs,based on which effective attack and defense strategies can be obtained.Currently,there exist many shortcomings in security researches on state estimation of CPSs,including but not limited to the following aspects:1)The existing researches on attack strategies of CPSs focused mainly on the attack on transmission channels and payed less attention to that on sensors;2)The majority of existing researches on attack strategies of CPSs took into account only one attack mode and a single sensor,while only few employed multiple modes to attack the multi-sensor CPSs at the same time;3)The existing defense strategies were able to resist the attack on the whole entities of the vector effectively,but were useless in partial attack on some entities;4)For most existing defense strategies for CPSs,the centralized fusion estimation algorithm with a unique fusion center were not able to cope with functional dysfunction of the fusion center.In response to these challenges,this thesis will study state estimation based attack and defense strategies of CPSs,including:For CPSs,two optimal FDI attack strategies are proposed under the constraint of energy at the attacker,which provide the optimal methods for the FDI attack on sensors and the FDI attack on both of sensors and transmission channels,respectively.First,the system model of attack is presented and two optimization problems are formulated.Then,the estimation error covariance of the CPS under the FDI attack is obtained by calculating the optimal Kalman gain.Next,by defining two positive semidefinite matrices,the optimization problems are transformed intoconvex problems and solved,followed by the optimal FDI attack strategies.Finally,simulation results verify the effectiveness of the proposed optimal attack strategies.Considering that sensors and transmission channels of a CPS are both attacked by FDI and denial-of-service(DoS),the optimal joint attack strategies are proposed under the constraint of energy at the attacker.Firstly,based on the FDI attack model and the DoS attack model,all optimization problems for all joint attacks of FDI and DoS are formulated.Then,the optimal joint attack strategies under all joint attacks are obtained by transforming and solving the optimization problems.Next,the influence of all joint attacks on the CPS is compared.Finally,the estimation error covariance is verified by simulations for the CPS using the optimal joint attack strategies.For FDI and DoS attacks on both of sensors and transmission channels of a CPS,a defense strategy based on a reference estimate is proposed.Then,the proposed defense strategy is improved and applied to the CPS over a wireless sensor network suffering from both FDI and DoS attacks,and the security estimate is obtained.Next,the performance of the proposed defense strategy based on a reference estimate is analyzed and its convergence is proved.Finally,the estimation performance of the defense strategies based on a reference estimate is simulated.The proposed optimal attack and defense strategies can enrich attack and defense theory of CPSs to a certain extent,and have a great theoretical and engineering value for resisting malicious attacks and improving security performance of CPSs.