A Hybrid Deep Learning based Network Anomaly Detection System with Real-time Dashboard for Zero Day Attack Mitigation

With the increasing sophistication and volume of cyber threats—particularly zeroday attacks traditional intrusion detection systems based on fixed rules or shallow learning models struggle to keep pace. This paper presents a Hybrid Deep Learning-Based Network Anomaly Detection System with Real-Time Dashboard for Zero-Day Attack Mitigation, designed to enhance detection accuracy, scalability, and real-time responsiveness. The proposed framework integrates Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks to automatically learn spatial-temporal patterns in network traffic. The system is evaluated using CICIDS2017 datasets, achieving an average accuracy of 97.88%, precision of 96.46%, and recall of 99.89%, even on previously unseen attack types. A Flaskbased web dashboard enables live monitoring of predictions with time stamped alerts, anomaly distribution charts. This work contributes a modular, interpretable, and high-performance anomaly detection framework that bridges the gap between advanced deep learning techniques and practical cyber security applications, especially in defending against zero-day exploits.