Consensus Control of Multiagent Systems Under DoS Attacks: A Dynamic-Key-Based Secure Scheme.

Area of Science:

• Control Systems Engineering
• Cybersecurity
• Artificial Intelligence

Background:

• Discrete-time multiagent systems (DTMASs) face vulnerabilities from denial-of-service (DoS) attacks.
• Secure state transmission is crucial for maintaining system integrity under cyber threats.
• Existing control methods may lack robustness against sophisticated DoS attacks.

Purpose of the Study:

• To investigate the secure consensus control problem for DTMASs under DoS attacks.
• To propose a novel model-free controller utilizing Q-learning (QL).
• To ensure robust consensus achievement despite communication disruptions and unauthorized access attempts.

Main Methods:

• A QL-based model-free controller is developed to autonomously synthesize control gain matrices.
• A dynamic encryption key strategy is implemented for secure state transmission in attack-free scenarios.
• A safe mode with dynamic parameter expansion is activated during DoS attacks to prevent quantizer saturation and maintain partial communication.

Main Results:

• Sufficient conditions for ensuring consensus under DoS attacks (frequency and duration) were derived.
• The proposed controller demonstrated significantly enhanced attack resistance in simulations.
• The method proved effective in securing state transmission through quantization-based encryption.

Conclusions:

• The QL-driven, model-free approach provides a robust solution for secure consensus control in DTMASs facing DoS attacks.
• The dynamic key and safe mode switching mechanism effectively mitigates attack impacts.
• Simulation results validate the superior performance and attack resilience of the proposed method compared to conventional techniques.