Deep Learning-Based Network Intrusion Detection and Prevention System

Current network intrusion detection systems struggle with feature representation, unknown attack detection, and coordinated response. This paper proposes an intelligent system that fuses NetFlow and payload features, employs a three-level detection engine (deep autoencoder, Transformer, GNN), and integrates with softwaredefined networking for real-time mitigation and adaptive feedback-driven model improvement. Experiments on a mixed dataset combining the CIC-IDS2018 and UNSW-NB15 show a detection rate of 98.7%, a false positive rate of 0.86%, and an average detection rate of 87.04% for unknown attacks, with real-time interception success reaching 99%.