Research On The Methods Of Data-Driven Based Cyber-Physical Systems Security Problems

With the continuous progress of science and technology,people's standards for actual demand are more abundant and more challenging.Cyber-Physical System(CPS)is a new generation of networked control systems with control attributes.Unlike the traditional control system,CPS integrates the physical process with the network environment for realizing the deep fusion and real-time interaction of the two processes in order to detect and control a physical entity in a safe,reliable,efficient and real-time manner.Because information interaction is in the network environment,it will inevitably be impacted by internal or external network malicious factors.How to deal with these threats is significantly important for CPS.Considering the nonlinearity or uncertainties of many actual CPSs and the corresponding models are often difficult to construct.Thus,it is more directly to solve series problems with huge data inside the CPSs.This paper dedicates to CPS security through data-driven technology for ensuring the reliability of the system.Based on the research background and research status elaborated in the first chapter and the fundamental theory cited in the second one,the following contents are mainly involved with the security control strategies analysis against random packet dropout and network attacks,respectively.The class of data dropout phenomenon in the cyber-physical systems,which is characterized as a specific statistical probability distribution model,its induced consequences on cyber-physical systems and the robustness to the dropout are given with theoretical derivations.The results extend the special case of Single-input-Single-output(SISO)nonlinear systems under dropout analyzed in previous literature,furthermore,a detection mechanism based the data is developed for identifying the time interval in the presence of data loss.At last,some simulated illustrations are presented to verify the feasibility of the mentioned strategies.Then a special network attack,namely,the intermittent Denial-of-Service attack and false data injection attack which let the value of output in specific time instants be zero,the corresponding security problem are analyzed,compared with the case of packet loss,DoS attacks are more uncertain and more malicious.The convergence of the control algorithm and the impacts on itself induced by DoS are proved with theoretical derivations.Through the difference values between the obtained output data and the desired expected tracking signals,it is simple to detect the specific time intervals when the process is under Denial-of-Service attacks.Moreover,a method based on the subspace identification predicted control is applied to restore the lost signals with a certain degree in those data loss periods.Finally,a simulation example of a nonlinear system is given to verify the validity of the proposed theory.At last,conclusions and further research directions are given in the fifth chapter.