A mechanistic fault detection and isolation approach using Kalman filter to improve the security of cyber physical systems

This paper seeks to understand and identify ways in which fault detection and isolation (FDI) methods can be utilized to enhance the cyber-security of cyber-physical systems (CPS). In this work, we have used state estimation to validate relations between process variables, termed invariants, thereby detect the onset of attacks. Multiple attack scenarios were studied and the proposed state estimation method was found to excel in detecting attacks conducted on sensors within the system; provided that not all sensors are compromised during the attack. However, the proposed methods for subsequent isolation and correction faced limitations due to delays or lack of information to pinpoint the attacked component. We demonstrated our proposed approach on a well-instrumented pilot scale water treatment plant equipped with controllers.