Benchmarking Machine Learning and Deep Learning Models for Web Attack Detection

The rapid evolution of web-based cyber threats, including SQL injection, cross-site scripting, and distributed denial-of-service attacks, has created a strong demand for intelligent and adaptive security solutions. Conventional detection mechanisms that rely on predefined rules are increasingly ineffective, as they struggle to recognize newly emerging and modified attack patterns. This study presents a comparative evaluation of multiple machine learning and deep learning techniques for the detection of web-based attacks. Supervised learning models such as Support Vector Machines, Decision Trees, and Random Forests are analyzed alongside deep learning architectures including Convolutional Neural Networks and Recurrent Neural Networks. Experimental analysis is conducted using standard benchmark datasets, namely CICIDS2017 and UNSW-NB15, to ensure fair and consistent evaluation. Model performance is assessed using metrics such as accuracy, precision, recall, and F1-score. The results reveal noticeable variations in detection capability across different attack categories, highlighting the strengths and limitations of individual models. The findings further demonstrate that combining multiple learning techniques can enhance detection effectiveness, indicating the potential of ensemble and hybrid approaches for building scalable and resilient web intrusion detection systems.