Peak Synergy Network: A Novel Marl Framework for Enhanced Cyber-Attack Detection

Cyber-attack detection is crucial for safeguarding sensitive data, digital assets, and critical
infrastructure. This research introduces the "Peak Synergy Network Model Framework
(PSNM)", a revolutionary method that improves detection accuracy, resilience, and
adaptability to evolving threats by utilizing Multi-Agent Reinforcement Learning (MARL).
Conventional MARL techniques face several drawbacks, including high computational
demands, the need for large volumes of data, and challenges in agent coordination. The
proposed framework addresses these issues by incorporating several innovative components.
First, comprehensive data preprocessing ensures high-quality training inputs through cleaning
and normalization. The intrusion response system is modelled as a multi-attribute decision￾making problem, optimizing time efficiency, cost, and resource constraints. The proposed
"Antecedent DQN Model" maximizes the predicted cumulative discounted reward and ensures
convergence to an optimal policy by employing Markov Decision Process parameters. A key
advancement is the "Shared Decision-Making Protocol," which enables agents to
independently learn policies through local observations, resulting in faster and more reliable
convergence with minimized communication overhead. Furthermore, "Premium
Hyperparameter Optimization" fine-tunes each parameter’s learning rate, accelerating training
and enhancing overall performance. The effectiveness of the proposed framework has been
validated through performance metrics, achieving 99.9% accuracy, 100% precision, 99.9%
recall, 99.9% F1-score, and a 0.0 false positive rate. These results confirm the framework’s
substantial potential for effectively detecting advanced and dynamic cyber threats.