A Hybrid Deep Learning based Intrusion Detection System

Abstract

The increasing dependence on digital communication networks has made the protection of information systems a significant challenge. As network environments continue to expand, cyber-attacks have become more frequent and diverse, reducing the effectiveness of conventional intrusion detection techniques that depend primarily on predefined attack signatures or manually designed rules. These traditional approaches often struggle to recognise evolving attack patterns and are less effective when dealing with complex network traffic. This project presents a hybrid deep learning framework for network intrusion detection by combining Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks. The CNN component is employed to learn meaningful feature representations from network traffic data, while the LSTM network captures sequential relationships that help distinguish normal behaviour from malicious activities. The integration of these two deep learning models enables the system to analyse both spatial and temporal characteristics of network traffic, leading to more reliable attack classification. The proposed framework is developed and evaluated using the CICIDS2017 dataset, which includes realistic network traffic containing both legitimate and malicious activities. The trained model is deployed through a Flask-based web application that supports user authentication, individual traffic prediction, batch prediction using CSV files, prediction history management, dashboard visualisation, and report generation. SQLite is used to maintain prediction records and application data efficiently. Experimental evaluation indicates that the hybrid CNN–LSTM model achieves strong classification performance across standard evaluation metrics, including accuracy, precision, recall, and F1-score. The combination of deep feature extraction and sequential learning enhances the system's capability to identify malicious network traffic more effectively than conventional machine learning approaches. The developed framework offers a practical and scalable solution for intelligent network intrusion detection and provides a foundation for improving cybersecurity in modern networked environments.