MODELS, SECURITY AND CONTROL OF CYBER-PHYSICAL SYSTEMS

Abstract

Modern control systems integrate physical processes with communication capabilities and computational resources. In general, the integration of the aforementioned capabilities improves system efficiency and operational performance, but at the same time introduces security concerns related to the intrusion of adversaries in the system. Moreover, the increasing amount of available sensor data poses new challenges in the task of monitoring malicious attacks against the system. During the last years, these motivations have led to the study of a particular class of control systems: the Cyber-Physical Systems (CPSs). Cyber-Physical Systems combine physical processes with computational resources in an interconnected framework, but expose control systems to new vulnerabilities and threats due to the inter-dependencies and links between cyber and physical layers. Examples of Cyber-Physical Systems include Supervisory Control And Data Acquisition (SCADA) systems, Power and Smart grids, where data fusion methodologies are useful for analyzing threats and faults. Within the cyber-physical security framework, Evidence Theory can be a powerful tool to help the control centers to make and plan decisions and/or countermeasures. In particular, in this thesis we develop a new approach for the diagnosis of faults and threats when cyber-attacks compromise physical operations in Power Grids (cyberphysical attacks). To handle the complexity of the fusion process and to minimize the computational overhead, we also propose a new way to model Cyber-Physical Systems in Evidence Theory framework. Moreover, through Graph Theory, risk assessment for Cyber-Physical Systems is rediscovered as an application field for Evidence Theory. We provide theoretical findings, supported by simulations results, able to manage risks arising from cyber-physical attacks. It is worth noticing that, CPSs act in dynamically changing environments and, despite significant advances in relevant areas, several challenges still hinder the development of high-assurance, robust and reconfigurable Cyber-Physical networks. Hence, in this thesis, we also address the problem of characterizing the robustness of Cyber Physical Systems, viewed as interconnected network systems, with respect to the interconnection structure. Specifically, we adopt the H2 norm, to measure the robustness of a CPS network against external disturbances. For networks arising from the composition of atomic structures, we provide a closed-form expression of the robustness, and we identify optimal composition rules. Furthermore, we also generalize the proposed model, using the class of M - matrices and their inverses. The problem of finding the optimal robust network structure has been analyzed as an optimization problem: we found several properties of the objective function and we also characterized the expression of the optimal solution.