Network Traffic Anomaly Detection using Deep Learning and Explainable AI

As networks become increasingly complex and interconnected, ensuring robust cybersecurity is paramount. Traditional rule-based systems, such as firewalls, struggle to detect sophisticated threats like zero-day attacks and data exfiltration, necessitating advanced detection methods. This study proposes a multi-class neural network model for network anomaly detection, achieving a high testing accuracy of 99.31% with a loss of 0.0537. To enhance model transparency and trust, we integrate Explainable AI (XAI) techniques, providing actionable insights into feature importance and decision-making processes. By combining deep learning’s pattern recognition capabilities with interpretability, our approach overcomes the limitations of conventional methods, offering a more effective and explainable solution for network security.