Learning-based minimum cost strategies for set reachability of Boolean control networks under data injection attacks

Area of Science:

• Control Systems Engineering
• Network Security
• Computational Biology

Background:

• Boolean control networks are crucial for modeling complex systems.
• False data injection attacks (FDIAs) pose significant threats to network security and reliability.
• Understanding attack vulnerabilities in set reachability is essential for robust system design.

Purpose of the Study:

• To investigate FDIAs with failure probabilities in Boolean control networks.
• To establish conditions for guaranteed set reachability under various attack and control scenarios.
• To develop efficient methods for finding optimal data injection sequences (DISs).

Main Methods:

• Derivation of necessary and sufficient conditions for set reachability under probabilistic FDIAs.
• Construction of a novel system for analyzing specific control input sequences (CISs).
• Development of an algorithm for minimum time DIS and a double deep Q network for large-scale networks.

Main Results:

• Established conditions for set reachability with probability one or above a given bound.
• Proposed a method to find minimum time DIS for specific CISs.
• Introduced a model-free approach using double deep Q network for cost-efficient DIS discovery in large networks.

Conclusions:

• The proposed methodologies effectively address FDIAs in Boolean control networks.
• The developed algorithms and methods are practical and efficient, demonstrated by biological examples.
• This research contributes to enhancing the security and reliability of control systems against sophisticated attacks.