Research on network anomaly traffic detection and intrusion prevention system driven by deep learning

Abstract

With the rapid development of cloud computing, the Internet of Things, and the Industrial Internet, the scale of network traffic continues to increase, and abnormal behaviors exhibit characteristics of concealment, variation, and persistence. Traditional detection methods have become unable to balance the accuracy of identification and the efficiency of coordinated defense. This paper focuses on the collaborative requirements of network anomaly traffic detection and intrusion prevention, builds a detection and defense system driven by deep learning, designs multi-granularity traffic feature representation and semantic encoding methods, proposes temporal-spatial correlation feature extraction, session context modeling, semantic constraint-enhanced discrimination, and a defense strategy generation mechanism driven by detection results. The experiments are based on 52,480 traffic session samples and 38-dimensional features. The accuracy, F1-score, and Defense Success Rate of the proposed method reach 95.4%, 94.8%, and 93.6% respectively, and the system stability score still reaches 90.9% under mixed perturbation conditions. The study shows that this method provides a technical reference for the intelligent construction of the cybersecurity protection system.