Attack concealment for cyber-physical systems using a mechanism borrowing from differential privacy

Area of Science:

• Cyber-physical systems security
• Discrete event systems
• Control theory

Background:

• Cyber-physical systems (CPS) integrate physical processes with computation, making them vulnerable to cyberattacks.
• Attackers can corrupt system observations, potentially evading detection by operators.
• State estimation is typically used by operators to identify specific attack types.

Purpose of the Study:

• To develop a cyberattack concealment mechanism for CPS.
• To ensure that system operators cannot detect attacks, even with knowledge of the defense mechanism.
• To protect CPS by making attack-induced observations indistinguishable from normal operations.

Main Methods:

• Modeling cyber-physical systems using finite automata within a discrete event systems framework.
• Introducing state sequence differential privacy to disguise observations.
• Designing a differential privacy mechanism to ensure approximate probability between attacked and random observations.

Main Results:

• The proposed state sequence differential privacy mechanism effectively conceals cyberattacks.
• Operators cannot detect the specific attack dictionary imposed on the system.
• The defense mechanism remains effective even when its details are public.

Conclusions:

• State sequence differential privacy offers a robust solution for cyberattack protection in CPS.
• This approach enhances the security of critical infrastructure, exemplified by a nuclear power facility case study.
• The method ensures attacker concealment by making malicious and benign system behaviors probabilistically similar to the observer.