Consensus-based Security Of Cyber-physical Systems Under DoS And Data Injection Attacks

In recent years,cyber-physical systems(CPSs)have received extensive attention and research with the development of sensing,communication,control and computation technology.CPSs realize the state perception,scientific decision-making and precise execution of automatic control systems by organically combining the information layer and the physical layer.However,CPSs face severe security challenges as the complexity of the system and the openness of the information layer communication protocol.As an important cyber-physical systems,multi-agent systems(MASs)have momentous applications in many key areas.On the other hand,due to the interaction between agents through the network and the opening control protocols,the stability and consensus of MASs may be lost if malicious attacks occur.Therefore,it has important theoretical and engineering significance to study the security and reliability of MASs under attacks.This thesis investigates the security of control systems under two typical attacks,i.e.,denial-of-service(Do S)attacks and false data injection attacks.The main work and contributions are summarized as follows.Firstly,the consensus problem of multi-agent systems under denial-of-service attacks is analyzed.State reconstruction mechanisms are designed for linear and nonlinear multi-agent systems,respectively.The system with the state reconstruction mechanisms can still achieve consensus when the observer network is subjected to a denialof-service attacks.Using the Lyapunov method,the criteria for ensuring consensus of MASs when the observer communication network and the controller communication network are both attacked is given.Secondly,the security of the consensus-based economic dispatch(ED)system under the injection attacks is studied.By analyzing the operation of the economic dispatch system under data injection attacks,the consistent convergence points of the system states are obtained.The virtual system based competitive interaction mechanism is utilized to eliminate or weaken the impact of the injection attacks on system consensus and make the convergence point optimal.Finally,the consensus of MASs under injection attacks is further investigated by combining the sensing-based state estimation mechanism.A sensing-based state estimation mechanism is provided.Using the discriminating mechanism to compare the state value obtained by perceptual estimation with the state information received by the communication network,MASs will determine which state values can be used for status updates,guaranteeing the security of MASs.