Abstract
A cyber-attack on a cyber-physical system (CPS) causes undesirable catastrophic failures. The secure water treatment (SWaT) facility is a CPS, prone to cyber-attacks. Therefore, an attempt has been made to detect faults based on physical failures. Cyber-attack data from the SCADA (Supervisory Control and Data Acquisition) unit of the SWaT system is used for CPS failure prognosis. Bayesian networks based on cyber-attack data predict the likelihood of failures and provide a probabilistic relationship between cyber-attack and physical failure. Assuming the attacker's intent space, various attack patterns are considered: (1) Single Stage Single Point (SSSP), (2) Single Stage Multi-Point (SSMP), (3) Multistage Single Point (MSSP), and (4) Multistage Multi-Point (MSMP) for six stages of SWaT-CPS. Bayesian Networks (BN) for different cyber-attack patterns provide an effective cybersecurity strategy for CPS. The new set of observed evidence (e.g., unexpected outcomes of cyber-attacks) can be updated in the BN Model.