Attack-defense Control And Secure Estimation For Cyber-physical Systems

Cyber-Physical Systems(CPSs),acting as the essential of the Fourth Industrial Revolution,closely integrate the physical plant with computation resource and perform the persistent coordination in essence.The core of CPSs,undergoing an ever-enriching cognition process,lie in the intelligent human-machine interaction systems with deep integration of control,computation,communication,cloud and perception.They oftentimes steer a higher degree of autonomy and collaboration.CPSs have been the novel crossdisciplinary carriers pushing forward an immense influence on modern real-world industrial applications,such as vehicular transportation,medical healthcare,smart grids and robotic systems.However,due to the intrinsically open attribute of their cyber realms,which in turn suffering from the vulnerability induced by information eavesdropping,privacy intrusion and communication interruption,the cyber threats could make the physical processes in the nature world susceptible by virtue of the cyber-physical integrations.From the perspectives of control and system engineering,security is the priority of CPSs,which tangibly leaves the rooms for exploring the corroborated secure defense control strategies against malicious corruptions while preserving the desired system performances.In this thesis,we consider both the denial-of-service corresponding to jamming the transmitted data flows leading to the timeliness violation and the false data injection attacks corresponding to impeding the transmitted data packets to destroy the trustworthiness.Allowing for the utilization of sliding mode control,robust control and optimization in conjunction with the analyses and syntheses for complex dynamical systems,our purposes boil down to the attack-resilient control deployments with specific concerns on the denial-of-service defense and secure estimation evolutions.The content of the thesis can be concisely recapped in what follows.1.In Chapter 2,for the control system whose controller-to-actuator channel suffering from intermittent denial-of-service attacks,we focus on the design of dissipativitybased sliding mode defense strategy.Based on the more applicable deterministic frequency and duration attack descriptions,the dynamical system under denial-ofservice can be viewed as the specific two-mode switching processes corresponding to attack-active and-silent,respectively.Then we meticulously derive the sufficient conditions on ensuring the stability in the complete time domain and establish the switched sample-data implementation.Besides,in view of the great importance of dissipativity for large-scale complex industrial processes,we subsume both the stability and dissipativity to the attack resilience characterizations.2.In Chapter 3,for the filtering process sufferers from intermittent denial-of-service attacks,our concerns lie in developing the attack-resilient robust filter.By virtue of the deterministic frequency and duration attack descriptions,the discrete-time filtering process under denial-of-service can be viewed as the specific two-mode switching processes,the current mode of which is determined by whether the attack is active or not.Then we put forward the sufficient conditions on the assurance of desired filtering performances in the complete time horizon as well as characterizing the attack resilience.Hence the filter parameters are determined by solving semidefinite programming.3.In Chapter 4,considering the control system susceptible to sparse actuator and sensor corruptions synchronously,our purpose lies in crafting the secure state estimation strategy in the light of switched counteraction-oriented brute force searches to meticulously ensure the convergence of state estimation as well as realize the autonomous decisions on the entry mode selections.Meanwhile,the performance degradation induced by malicious attacks can be predominantly mitigated.Besides,in the sense that the curse of dimensionality is inevitable for the combinational searches,we hence discuss the potential applicability of set cover intensified secure estimation to certificate the estimation effectiveness and reduce the number of searching candidates.4.In Chapter 5,for the control system suffers from sparse actuator and sensor misappropriation synchronously,we show our specific concerns on establishing the illuminated secure estimation schemes with the aid of switched-counteraction descriptor full-and reduced-order sliding mode observers,respectively.Based on the novel adaptive online searching logic,the sufficient conditions ensuring the convergence of real-time state and attack reconstructions and realizing the autonomous decisions on entry mode location can be well derived,which leads to determining the observer gains in terms of convex programming.Owing to the attack reconstructions,these strategies can manifestly back up the attack detection and identification in essence.5.In Chapter 6,for the distributed control systems suffering from sparse actuator and sensor attacks,we particularly focus on designing the Distributed Projected Heavy-Ball Estimator and Distributed Projected Luenberger Observer algorithms.Based on the optimization theory,the alternations between multiple subgradient descents and a projection step(in conjunction with)time-window onward update can corroborate the attenuation of estimation cost leading to the estimation convergence.Besides,the intensified Combinational Vote Locations strategy backs up the autonomous decision on entry mode selection and the distributed implementation of projection operator.The designed schemes render the estimation effectiveness while potentially speeding up the convergence.