ABSTRACT
Nowadays, arming automation systems with modern IT results in a number of benefits, such as low maintenance and installation costs, and increased system interoperability. However, on the other hand, the exposure to public networks suggests that the control network is not simple and isolated anymore and can be easily compromised by malicious attacks; see, e.g., [63], [11], and [194]. In this chapter, the term ’resilience’, standing for minimizing the impact of the adversary, is captured in both the physical layer and the cyber layer. In the physical layer, the adversary usually takes the form of external disturbance, and resilience is actually the robustness of the system to such disturbances. Thus, we introduce H∞ optimal control to resist external disturbance in the physical layer. In the cyber layer, we restrict the adversary to be DoS attacks, and resilience can be seen as defending against such attacks. We use IDSs to detect the anomaly in the cyber layer, such as overlong time delays. Then, IDSs raise alarms so that malicious attacks can be removed automatically. Thus, IDSs are used to defend against malicious behavior in the cyber layer. In a nutshell, resilient control suggests a coupled design of H∞ optimal control and IDSs configuration for CPSs, and such a layered design can be seen in Figure 13.1
Contributions of this chapter are summarized as follows: firstly, a hybrid model for RCS is proposed, in which the cyber security part is modeled as a stochastic Markov game and evolves on a different time scale from that of the physical control part. Two types of discrete operators, the delta operator and forward shift operator, are used in the hybrid model according to their respective numerical advantages. Secondly, the joint optimality is proposed to capture the interdependence of the physical and cyber parts of the hybrid model. And last but not least, a coupled design methodology is proposed to achieve the goal of joint optimality. The algorithms of the coupled design methodology are unified into the framework of ILMIs.
