Intrusion Detection System for Web Applications Using CNN–LSTM Hybrid Deep Learning Model

Abstract ―​

This paper presents an advanced hybrid deep learning-based Intrusion Detection System (IDS) for web applications using a combination of Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks. The proposed model leverages CNN for spatial feature extraction and LSTM for capturing temporal dependencies in sequential network traffic data. With the rapid increase in sophisticated cyberattacks such as SQL Injection (SQLi), Cross-Site Scripting (XSS), and Distributed Denial of Service (DDoS), traditional signature-based and rule-based IDS systems fail to provide adequate protection, especially against zero-day attacks. The experimental results demonstrate that the proposed CNN–LSTM model achieves an accuracy of 97.37%, a recall of 98.42%, and a low false positive rate of 2.88%, making it highly reliable for real-time intrusion detection. The system is scalable, efficient, and capable of adapting to evolving cyber threats, making it suitable for deployment in modern web-based environments.