MWAD: A Multimodal Web Attack Dataset for AI-Driven SQL Injection Detection

The Multimodal Web Attack Dataset (MWAD) is a novel labelled cybersecurity dataset developed as part of doctoral research in cybersecurity at La Trobe University, Australia. Unlike conventional web attack datasets that provide either application-layer HTTP requests or network traffic independently, MWAD integrates HTTP request data with their corresponding network-flow features, enabling multimodal analysis of web application attacks. MWAD was developed to support research into artificial intelligence (AI) and machine learning (ML) for cybersecurity, particularly in the detection of web application attacks. By providing complementary application-layer and network-layer information within a unified dataset, MWAD enables researchers to investigate multimodal learning approaches that can improve the accuracy and robustness of AI-driven web attack detection systems. The dataset is intended to facilitate reproducible research, benchmark the performance of emerging AI models, and contribute to advancing the development of intelligent cybersecurity solutions. The dataset generation methodology, design, and description are presented in the peer-reviewed journal article: Yeboah, P. N., et al. (2026). SQL injection detection using self-supervised pre-training and multimodal techniques. Intelligent Systems with Applications, Elsevier, 31, 200702. https://doi.org/10.1016/j.iswa.2026.200702 The dataset contains 1,000 labelled samples (500 benign and 500 SQL injection attacks) generated in a controlled experimental environment using Metasploitable, SQLMap, Wireshark, Tcpdump, and NFStream. MWAD is intended to support research on: Artificial intelligence and machine learning for cybersecurity SQL injection detection Multimodal learning Deep learning and transformer-based models Intrusion detection systems Feature-level and decision-level data fusion This dataset underpins research published in Intelligent Systems with Applications (Elsevier). The GitHub repository contains comprehensive documentation, the dataset generation methodology, and usage instructions.