Adaptive neural network control for Markov jumping systems against deception attacks.

Area of Science:

• Control Systems Engineering
• Network Security
• Artificial Intelligence

Background:

• Deception attacks pose significant threats to the stability and reliability of Markov jumping systems.
• Existing monitoring mechanisms often struggle with dual-mode complexities and unbounded false signals.
• Ensuring system resilience requires advanced control strategies capable of adapting to dynamic changes and adversarial inputs.

Purpose of the Study:

• To propose an adaptive neural network control strategy for mitigating deception attacks in Markov jumping systems.
• To develop a unified Markov chain model integrating system and actuator states.
• To design an asynchronous control law for robust system performance under attack.

Main Methods:

• Utilizing two independent Markov chains to model system and actuator states, then merging them into a joint Markov chain.
• Employing an adaptive neural network to approximate unbounded false signals from deception attacks.
• Implementing a mode monitoring scheme for an asynchronous control law linking joint Markov chain modes with the controller.

Main Results:

• Sufficient criteria for mean-square bounded stability derived using Lyapunov theories.
• The proposed adaptive neural network control effectively approximates unbounded false signals.
• The asynchronous control law successfully links mode information with fewer controller modes.

Conclusions:

• The developed adaptive neural network control strategy effectively mitigates deception attacks in Markov jumping systems.
• The unified joint Markov chain and mode monitoring scheme enhance system robustness.
• Numerical experiments validate the proposed method's effectiveness in ensuring system stability.