Research On The Security Control For Cyber Physical Systems Under Multiple Attacks

Cyber-Physical Systems(CPSs)are a kind of novel intelligent complex network system that integrates physical perception,calculation,communication and control to realize real-time perception and distributed control.With the rapid development of communication technology,CPSs have also developed booming and play an increasing important role in many areas,such as key infrastructure and government agencies.Up to now,CPSs have successfully applied in a variety of areas,such as smart grid,smart city,smart medical,smart transportation,and so on.It is worth noting that these applications must be built on the basis of ensuring system security,otherwise it may bring serious threats to people's lives and properties.However,in practice,the sensing devices of CPSs are often distributed with the logic and geographic location.Although distributed collaboration improves the performance of the system,it also increases the vulnerability.Therefore,it is of practical significance and theoretical value to investigate the security problem of CPSs.The current research on CPSs security issue has received much attention,and some results have presenteed.Most of the exisiting results focus on the security issues of CPSs from an information perspective.Sine the CPSs pose the characteristics of deep integration of network space and physical space,it is necessary to investigate the security control from the control perspective to guarantee security over the whole CPSs.Although some progress has been made in the study of CPSs security issues from the control perspective,there are still many open challenging problems deserving further investigation.For example,most of existing results focus on one type of cyber attack and established an invariable attack model.There is a lack of research on sophisticated attack type and a modeling method for fusing various cyber attacks dynamic evolution rules;some existing centralized security control method is difficult to deal with practical application scenarios where the distributed control system((such as smart grid systems,unmanned aerial systems,etc.))are attacked by cyber attacks;in the complex network environment,CPSs inevitably faces many irrational factors such as external disturbance,uncertainty,ctuator saturation,and so on.These uncertainty factors may make the distributed security controller design process and security analysis becoming difficult.In view of the deficiencies,this thesis focus on studying the distributed security control problem for multi-agent CPSs under multiple attacks.The specific research contents are as follows:1)The problem of security control for multi-agent CPS under false data injection attack is studied,where the information channels and control signal transmission channel are attacked by false data injection attacks simultaneously.Based on the analysis of the process of false injection attack,the Bernoulli process is applied to model randomly occurring false data injection attacks.To mitigate the false data injection attack impact on the system performance,a distributed security controller is designed.Furthermore,an event triggering control scheme over a finite-horizon is first proposed,which can reduce the can reduce the limited energy consumption of the transmission channel,ensure the control performance,and improve the robustness of the system.2)The problem of secure consensus control for time-varying multi-agent CPS under disturbance and joint attacks is studied.A cunning attacker simultaneously launches fake data injection attacks and denial of service attacks on the forward and the feedback control channels of the feedback control system.On the basis of stochastic characteristics of joint attacks,the mutually independent Bernoulli process is used to construct reasonable mathematical models for randomly occurring joint attacks.A distributed security consistency control strategy is proposed to ensure the security of the system.3)The problem of security control for time-varying multi-agent CPSs under actuator saturation and hybrid attacks is considered.It is the first attempt to construct an appropriate mathematical model by introducing random variables that obey Markov processes to describe hybrid attacks.Then,to reduce the impact of attacks on system performance,the switching controller parameters are designed by fusing dynamic evolution law of hybrid attacks.Finally,the random process theory is applied to analyze the influence of single control strategy and switching control strategy designed under hybrid attack on control system performance.4)The issue of security control for unmanned aerial systems with random switching topology and replay attacks is considered.A Markov random process is introduced to describe unmanned aerial systems topology changes in real time position.Due to the attacker launched replay attacks,it is difficult to measure whole states,so an observer is designed.In order to ensure the security of the unmanned aerial systems,based on observer secure consensus control scheme is designed.To verify the practicability and feasibility of the proposed security control method,the unmanned aerial systems experimental platform is applied.