Advancing Cyber-Physical Systems: Enhancing Network Security and Fully Distributed Estimation and Control

The dynamic interplay of socio-economic changes and technological progress has driven industries to embrace advanced methodologies for competitiveness and ecological responsibility. Above all, the evolution of cyber-physical systems (CPSs) has revolutionized industrial automation through the integration of cyberspace and the physical world. While CPSs management models have evolved from monolithic to distributed and intelligent paradigms, challenges persist in modeling the complexity and communication requirements inherent in this decentralization. Focusing on the main needs related to the coordination and organization of agents within distributed networks, this Thesis seeks to offer a thorough exploration of innovative concepts and open challenges in the realm of cyber-physical systems, multi-agent systems, and distributed algorithms. In particular, two pivotal and notably sensitive facets in the context of CPSs are explored: enhancing cybersecurity and agent privacy aspects, and achieving state awareness for system monitoring. This Thesis introduces innovative solutions to these crucial challenges. With regard to the first objective, although real-time monitoring of processes is mandatory for implementing innovative services and increasing efficiency, one of the main problems that distributed frameworks have to face is their exposure on the network, which makes them prone to cyber incursions by malicious third parties. Guaranteeing privacy and resistance against eavesdroppers while allowing agents to reach an agreement on some shared variables is an essential feature to foster the adoption of distributed protocols. To address these needs, this Thesis presents innovative approaches based on geometric coordinate shift and on the exploitation of nonlinear or chaotic dynamics that can be easily integrated into distributed networks to improve privacy and eavesdropping resistance, without compromising computational efficiency, speed, or bandwidth. Moreover, the Thesis includes the application of the geometric-based cryptographic methodology to address a security issue within the GOOSE protocol for digital substations. Lastly, the problem of identifying and reacting to false data injection attacks in distributed systems is analyzed, proposing innovative synchronous and asynchronous approaches that do not rely on stringent assumptions about the topology or the attacker. Concerning the state awareness, the motivation stems from the fact that, in the realm of multi-agent systems, distributed estimation, and control algorithms are the key to achieving a collective understanding of dynamical systems, while facing challenges such as limited information, communication problems, and adaptation to dynamic environments. To date, the greatest open challenge to be addressed in this context pertains to the distributed nature of the estimation mechanism, as agents are assumed to be able to directly measure only a subset of the entire system, which is characterized by strong interdependent relationships. To achieve this goal, in this Thesis a new methodology based on distributed estimation and control is proposed that successfully overcomes the limitations of previous state-of-the-art work. Specifically, the approach includes a novel and fully distributed mechanism for the design of finite-time estimation and control gains that allows the system to be self-configuring and resilient to node failures. Moreover, taking a further step beyond the traditional concept of distributed estimation, which considers the system as a single monolithic entity to be collectively estimated, the Thesis presents a new paradigm for distributed estimation and control based on individual interdependent subsystems, applied to both nonlinear and linear systems. The cornerstone of this strategy is the exploitation of a novel property of linear and time-invariant systems, namely negativizability. To conclude, the Thesis examines two pivotal aspects concerning the evolution of CPSs within the contemporary landscape of technological advancement. In the first part, the robustness of such systems to possible cyber-attacks by external or even legitimate entities is addressed, proposing solutions to strengthen the privacy and security of the entire network, while maintaining the characteristics of rapidity and simplicity of implementation. In the second part, innovative approaches to achieve situational awareness of the system are presented, with distributed estimation and control strategies for systems characterized by both linear and nonlinear interdependencies.