A CircleSA-Optimised feature selection and FNP-SE deep learning framework for enhanced network intrusion detection in enterprise systems
Brij B. Gupta, Shin-Hung Pan, Akshat Gaurav, Varsha Arya, Razaz Waheeb Attar, Amal Hassan Alhazmi, Ahmed Alhomoud, Kwok Tai Chui · Enterprise Information Systems · 2025
Modern communication networks must be kept against the development of cyber threats by using network intrusion detection systems in enterprise systems. In this context, this paper presents an AI-driven CircleSA-Optimised FNP-SE Framework. The model selects the optimal features by using the Circle Search Algorithm (CircleSA), therefore decreasing computing overhead while maintaining important spatial and temporal properties. The system enhances feature recalibration by combining squeeze-and-excitation (SE) blocks with feature pyramid networks (FNP), hence capturing hierarchical relationships. With a 97.86% accuracy, precision of 97.96%, recall of 97.86%, and an F1-score of 97.86%, experimental assessments on benchmark datasets show extraordinary performance.