Detecting Web Attacks with End-to-End Deep Learning

The rapid proliferation of web-based technologies has heightened the risk and complexity of cyber threats, making traditional detection mechanisms increasingly inadequate. This research introduces an end-to-end deep learning approach to detect and classify web attacks with improved precision and adaptability. The proposed model integrates AutoEncoder for anomaly detection and Long Short-Term Memory (LSTM) networks for sequential pattern analysis, enabling robust identification of malicious behaviors within network traffic. Comparative performance evaluation with conventional algorithms such as Support Vector Machine (SVM) and Naïve Bayes demonstrates significant improvements, with the proposed system achieving up to 81.99% accuracy and outperforming others in key metrics such as precision, recall, and F1-score. The results indicate that the deep learning-based framework can serve as a scalable and intelligent solution for modern cybersecurity infrastructure, offering real-time threat detection with minimal false positives.