Research On Secure State Estimation For Cyber-Physical Systems Subject To Malicious Attacks

In recent years,a new generation of network information technology characterized by high-speed interconnection,intelligence and convenience,and comprehensive integration is developing vigorously,promoting the transformation of human society from industrial society to information society.Cyber-physical System(CPS)is a type of networked system that integrates "3C"(Computation,Communication & Control)technologies.Through organic integration and deep collaboration of dynamic information,CPS has been widely used in smart grid,environmental monitoring,transportation,medical diagnosis and other fields through the organic fusion of dynamic information and deep collaboration.The interconnectivity of CPS components and the openness of the network environment give potential malicious attackers the opportunity to hijack and tamper with packets,making it more challenging to maintain the secure of CPS network systems.Governments have taken a series of defensive measures and formulated relevant laws and regulations to secure CPS networks and data,which have attracted great attention from all walks of life.How to improve the stability and reliability of CPS in a non-secure network environment has become an important research topic in academia,among which,accurate state estimation is a key part of securing CPS.In the CPS state estimation problem for non-ideal network transmission environment,the data processing center can estimate the true value of the internal state of the physical system based on multi-source sensory data and interaction information.Based on the estimated data provided,decision makers can make scientific judgments and decisions and take reliable countermeasures to ensure the data security and stable operation of the physical system.Much progress has been made on the problem of CPS secure state estimation.However,most of the research results focus on the secure state estimation of a single channel or performance analysis based on a single attack type,while there is a lack of research on the problem of secure state estimation under multiple locations and diverse types of attacks.Based on the above deficiencies,this thesis establishes a secure state estimation model based on multi-aware information for insecure network information transmission environment,applies the mean residence time theory,theory and Lyapunov stability criterion to design safe and reliable centralized estimators and distributed estimators,and focuses on the distributed secure state estimation algorithm when the attack location is diversified and the type is complicated.The main work is as follows.(1)The finite-time secure state estimation problem when the transmission channel of the measurement information is attacked by false data injection is studied for the centralized estimation structure.A centralized secure state estimation error system model is developed by describing the sensed data after the attack into two parts that are correlated and uncorrelated with the original measurement information.Based on the mean residence time theory and Lyapunov stability criterion,sufficient conditions are derived for the finite-time stability of the secure state estimation error system with a given performance,and the algorithmic weights of the secure state estimator satisfying the conditions are designed.Then,the design problem of finite-time safety state estimator algorithm weights is transformed into a set of linear matrix inequality solution problems.Finally,the effectiveness of the finite-time safety state estimator design algorithm is verified by practical cases.(2)The problem of secure state estimation under a two-channel hybrid attack is studied for a distributed estimation structure.A novel system for distributed secure state estimation consisting of nine subsystems is established by describing the post-attack measurement and neighborhood information as two parts that are correlated and uncorrelated with the original measurement and neighborhood information.Based on the mean residence time theory and Lyapunov stability criterion,sufficient conditions are derived for the secure state estimation error system to be exponentially stable and have a given performance,and the weights of the distributed secure estimation algorithm satisfying the conditions are designed.Then,the design problem of the safety state estimator is transformed into solving a set of linear matrix inequalities.Finally,the effectiveness of the proposed algorithm is verified by a case study.The results of this thesis can,on the one hand,provide theoretical guidance and expand new research ideas for CPS security under malicious data attacks.On the other hand,it can provide accurate data support for the control,decision and management of CPS,and guarantee the safe and stable operation of the system.