AI-ENHANCED INTRUSION DETECTION SYSTEM USING HYBRID MACHINE LEARNING AND DEEP LEARNING TECHNIQUES

This paper presents an experimental evaluation of an AI-enhanced Intrusion Detection System (IDS) that integrates classical machine learning, deep learning, and Natural Language Processing (NLP)-driven log analysis.Using the NSL-KDD and CICIDS2017 datasets, the study investigates detection accuracy, false-positive reduction, and real-time classification capabilities.The hybrid IDS combines Random Forest, Autoencoderbased anomaly detection, and BERT-based log embedding for semantic threat understanding.Results show significant improvement over traditional IDS models: a 17.4% increase in detection accuracy, a 24.6% reduction in false positives, and a 19.3% improvement in F1-score.Explainable AI (XAI) modules further enhanced analyst trust by 32%.Findings suggest that AI-driven hybrid IDS models offer robust, interpretable, and scalable detection in modern network environments.