Security control for networked control systems with deception attacks: A stochastic model predictive control approach

Area of Science:

• Control Engineering
• Cybersecurity
• Optimization Theory

Background:

• Networked control systems are vulnerable to sophisticated deception attacks.
• Existing methods often struggle with unknown-bounded stochastic attack signals.

Purpose of the Study:

• To design an innovative security control methodology for linear networked control systems (LNCS) under stochastic deception attacks.
• To mitigate conservatism in security control design by employing chance constraints.

Main Methods:

• Introduced an ambiguity set to characterize potential deception signals using mean-covariance constraints.
• Formulated chance constraints on system state and control variables.
• Applied distributionally robust optimization for a deterministic convex reformulation.
• Utilized stochastic model predictive control (SMPC) for recursive feasibility and convergence.

Main Results:

• Successfully addressed unknown-bounded stochastic deception attacks.
• Mitigated conservatism in security control design.
• Demonstrated recursive feasibility and convergence of the controlled system.
• Validated the approach's superiority through simulations on a DC-DC boost converter.

Conclusions:

• The proposed security control methodology effectively enhances the resilience of LNCS against stochastic deception attacks.
• The integration of ambiguity sets, chance constraints, and SMPC offers a promising direction for robust control system design.