Cyber–Physical Systems are the technological backbone of modern industrial, energy, and transportation infrastructures. They integrate computation, communication, and physical processes to enable automation, real-time control, and large-scale coordination. Yet, this tight interconnection also introduces fragility: faults or cyber attacks in the digital layer can quickly propagate through physical feedback loops, compromising safety and reliability. Ensuring the security and resilience of such systems therefore requires a unified approach, where estimation, control, communication, and interdependence are treated as mutually reinforcing aspects of the same problem. This thesis develops a comprehensive framework for model-based security and monitoring in resilient control systems. The central idea is that resilience begins with trustworthy perception — the ability of a system to maintain an accurate representation of its physical state even in the presence of deception or uncertainty. To this end, hybrid estimation architectures are designed to detect and isolate faults or intrusions in real time. By combining stochastic filtering with deterministic reasoning, these observers achieve a balance between sensitivity to attacks and robustness against modeling errors, allowing for consistent reconstruction of compromised signals. Building on this foundation, the research investigates how the reliability of information changes over time in adversarial environments. It introduces a formal metric to quantify how long a system operates with incorrect or outdated information, capturing the dynamics of misinformation caused by attacks such as data falsification or denial of service. This concept is analyzed within a game-theoretic framework that models the strategic interaction between defenders and adversaries, providing analytical insight into optimal attack and protection strategies. The work then extends these ideas to distributed and cooperative settings. In networked systems, agents must coordinate their actions through potentially unreliable communication channels. To address this challenge, the thesis proposes a secure consensus mechanism that integrates trust assessment directly into the coordination process. Each agent evaluates the reliability of its neighbors and adjusts its interactions accordingly, enabling agreement and stability even in the presence of malicious participants or uncertain data. The same principles are also adapted to ensure privacy-preserving coordination, allowing agents to cooperate without disclosing sensitive internal information. Finally, the thesis scales this perspective up to the systemic level, focusing on the interdependencies among critical infrastructures such as energy, telecommunications, transport, and finance. Using an agent-based simulation framework, it models how local cyber incidents can trigger cascading effects across sectors and how uncertainty — both statistical and epistemic — influences resilience. This modeling approach provides a tool for national-scale risk assessment and supports the design of coordinated protection strategies for interconnected infrastructures. Overall, the thesis presents an integrated view of security and resilience in cyber–physical systems, linking low-level estimation and detection to high-level systemic analysis. By connecting mathematical modeling, information dynamics, and cooperative control, it contributes to the development of infrastructures that are not only intelligent and interconnected but also robust and trustworthy in the face of evolving cyber threats.
Model based security and monitoring systems for resilient industrial control systems: A systematic approach to Cyber Physical Systems Security / Bonagura, V.. - ELETTRONICO. - (2026).
Model based security and monitoring systems for resilient industrial control systems: A systematic approach to Cyber Physical Systems Security
BONAGURA, VALERIA
2026
Abstract
Cyber–Physical Systems are the technological backbone of modern industrial, energy, and transportation infrastructures. They integrate computation, communication, and physical processes to enable automation, real-time control, and large-scale coordination. Yet, this tight interconnection also introduces fragility: faults or cyber attacks in the digital layer can quickly propagate through physical feedback loops, compromising safety and reliability. Ensuring the security and resilience of such systems therefore requires a unified approach, where estimation, control, communication, and interdependence are treated as mutually reinforcing aspects of the same problem. This thesis develops a comprehensive framework for model-based security and monitoring in resilient control systems. The central idea is that resilience begins with trustworthy perception — the ability of a system to maintain an accurate representation of its physical state even in the presence of deception or uncertainty. To this end, hybrid estimation architectures are designed to detect and isolate faults or intrusions in real time. By combining stochastic filtering with deterministic reasoning, these observers achieve a balance between sensitivity to attacks and robustness against modeling errors, allowing for consistent reconstruction of compromised signals. Building on this foundation, the research investigates how the reliability of information changes over time in adversarial environments. It introduces a formal metric to quantify how long a system operates with incorrect or outdated information, capturing the dynamics of misinformation caused by attacks such as data falsification or denial of service. This concept is analyzed within a game-theoretic framework that models the strategic interaction between defenders and adversaries, providing analytical insight into optimal attack and protection strategies. The work then extends these ideas to distributed and cooperative settings. In networked systems, agents must coordinate their actions through potentially unreliable communication channels. To address this challenge, the thesis proposes a secure consensus mechanism that integrates trust assessment directly into the coordination process. Each agent evaluates the reliability of its neighbors and adjusts its interactions accordingly, enabling agreement and stability even in the presence of malicious participants or uncertain data. The same principles are also adapted to ensure privacy-preserving coordination, allowing agents to cooperate without disclosing sensitive internal information. Finally, the thesis scales this perspective up to the systemic level, focusing on the interdependencies among critical infrastructures such as energy, telecommunications, transport, and finance. Using an agent-based simulation framework, it models how local cyber incidents can trigger cascading effects across sectors and how uncertainty — both statistical and epistemic — influences resilience. This modeling approach provides a tool for national-scale risk assessment and supports the design of coordinated protection strategies for interconnected infrastructures. Overall, the thesis presents an integrated view of security and resilience in cyber–physical systems, linking low-level estimation and detection to high-level systemic analysis. By connecting mathematical modeling, information dynamics, and cooperative control, it contributes to the development of infrastructures that are not only intelligent and interconnected but also robust and trustworthy in the face of evolving cyber threats.| File | Dimensione | Formato | |
|---|---|---|---|
|
Bonagura__DAuSy_38cycle_Final_Thesis_signed2.pdf
Open Access dal 10/11/2026
Tipologia:
Tesi di dottorato
Licenza:
Tutti i diritti riservati
Dimensione
35.48 MB
Formato
Adobe PDF
|
35.48 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.


