Research On State Estimation And Biasing Injection Attack Detection In Cyber-physical Systems

Cyber-physical systems(CPS)enable the close integration of Physical processes and network infrastructure with the help of ubiquitous computing resources and communication capabilities.CPS has penetrated into modern society and is widely used in various fields such as energy,transportation,advanced manufacturing and medical and health care.The security of CPS against sensor network attacks has always been the focus of attention.However,due to the tight integration of network and physical components,CPS has extensible vulnerabilities that surpass traditional network systems.Complex and malicious network attacks continue to appear,affecting the operation of CPS,resulting in performance degradation,business interruption,system failure.Security state estimation and attack detection and identification technology are the key technologies to ensure the reliable monitoring and operation of critical security CPS.In this paper,we study the detection and state estimation of biasing attack in cyber-physical systems,propose the algorithm of biasing detection and state estimation in the single sensor network and distributed sensor network,and realize the detection of biasing attack through the combination of Kalman filter and T detection method.As a special kind of false data injection attack(FDI),biasing injection attack(BIA)infuses constant false bias data by attacking the operating data in the cyber-physical system,which will lead to the error of system state estimation and affect the normal operation of the system.Therefore,the optimal security state estimation and attack detection of CPS under BIA are studied in this paper.The security state estimation algorithm and specific biasing attack implementation strategy based on different CPS system model structures are proposed.Two different scenarios of biasing attack detection are discussed,which can be divided into centralized single-node sensor network and distributed multi-sensor network.Under the single node network(SNN),based on the principle of minimum trace of the kalman filter state estimator and a detector with the combination of attack detection model we get system observation object on the analysis of the same attack scenario based on the optimal state estimation system of measuring residual change.By taking the target observation function of deviation into consideration to judge the system is under attack or not,we further come up with the target function of attack detection threshold and the T test precept based on hypothesis test.The final matlab simulation results make it clear that the proposed method can effectively detect BIAs in a certain short time,and the detection rate is more than 2% higher than that of traditional detection methods,furthermore,it has stronger robustness.Distributed multi-sensor network(MSN)is composed of multiple centralized single-node networks.On the basis of realizing the optimal state estimation of single node sensor,the attack model was reconstructed,the CPS comprehensive security state estimation algorithm based on data fusion was proposed,and the weight of observation data in each region was reasonably allocated.Then,the multi-region correlation analysis was carried out on the system measurement residual,and the attack detection judgment rules were summarized.Finally,a vehicle system with multi-sensor networks is taken as an example to illustrate the applicability and stability of the proposed security state estimation,and the efficiency of attack detection is more than 3% higher than that of single node.