Multistage adaptive cyberattack in power systems with CNN identification feedback loops.

Area of Science:

• Cybersecurity
• Energy Systems Engineering
• Artificial Intelligence

Background:

• Digital integration in hybrid hydrogen-power networks creates new cybersecurity vulnerabilities.
• Existing cyberattack models are static and lack dynamic adaptability, hindering realistic defense strategies.

Purpose of the Study:

• To propose a novel, three-stage, dynamically evolving cyberattack framework (CDB-TAS) for hybrid hydrogen-electric networks.
• To enhance cyberattack realism and effectiveness by incorporating dynamic adaptation, multi-stage coordination, and obfuscation.

Main Methods:

• Developed a Cyberattack Design Based on CNN-DQN-Blockchain Technology for Targeted Adaptive Strategy (CDB-TAS).
• Utilized Convolutional Neural Network (CNN) for reconnaissance and anomaly detection.
• Employed Double Deep Q-Network (Double DQN) for dynamic strategy refinement.
• Integrated a private blockchain for attacker-side obfuscation and decentralized coordination.

Main Results:

• CDB-TAS induced up to 15% voltage drop at critical buses and disrupted over 600 MW of load.
• Achieved 23.4% higher disruption efficiency compared to baseline attacks.
• Demonstrated lower anomaly detection rates due to continuous feedback adaptation and obfuscation.

Conclusions:

• The study presents the first integrated framework combining CNN, reinforcement learning, and blockchain from an adversarial perspective.
• CDB-TAS offers new insights into the evolving threat landscape of hybrid hydrogen-power systems.
• Findings guide the development of future cyber-resilience strategies for multi-energy systems.