Abstract

The security of Cyber-Physical Systems (CPS) has been the center of attention in the past decades. Developing methodologies to detect or estimate cyber-attacks on sensor measurements
and actuator inputs is essential for ensuring the safe and reliable performance of these interconnected systems of systems. Considering the stealthy nature of cyber-attacks, combined with potential faults, additional challenges emerge, which this thesis addresses through the lens of control theory.

In control theory, several methodologies have addressed the decoupling of unknown inputs, such as faults and disturbances. However, the simultaneous presence of faults and cyber-attacks
presents challenges that are not fully developed in the context of CPS. The first part of this thesis proposes a methodology, consisting of the construction of two plant-side monitoring filters to detect and isolate faults and cyber-attacks including covert and zero dynamic attacks. The findings are supported through analytical and simulation studies.

As the second challenge, a multi-rate approach is employed to estimate actuator cyber-attacks in the presence of sensors and actuators faults in CPS. A plant-side fault monitoring filter is
augmented with the physical system, and its residual, along with the plant’s outputs, is sent to C&C with a specified mechanism. Cyber-attacks are isolated from faults through the received
information in the C&C and a secondary observer. Consequently, a delayed Unknown Input Observer (UIO) is constructed to estimate the actuator cyber-attacks. The effectiveness of the proposed
methodology is evaluated through numerical case studies.