A Lightweight Deep Learning Model for Real-Time Anomaly Detection in Network Traffic

Abstract

This paper introduces a computationally efficient deep learning framework for real-time anomaly detection in network traffic environments characterized by high volume and dynamic behavior. Unlike conventional intrusion detection systems that rely on static signatures or resource-intensive architectures, the proposed model integrates a compact Convolutional Neural Network (CNN) with a streamlined Long Short-Term Memory (LSTM) module to jointly capture spatial and temporal characteristics of network flows. The model is specifically optimized for low-latency inference, making it suitable for deployment in resource-constrained environments such as IoT and edge networks. Experimental validation on the UNSW-NB15 and CIC-IDS2017 datasets demonstrates that the proposed approach achieves an accuracy of 97.8% while maintaining a significantly reduced false positive rate and computational overhead. The results indicate that the proposed architecture effectively balances detection performance and efficiency, offering a practical and scalable solution for modern cybersecurity systems requiring real-time responsiveness.