Real-time threat detection of phishing attacks from URLs using an intrusion detection system (IDS) with a character-level convolutional neural network (Char-CNN)

Avila-Gutierrez C.; Mejia-Giron J.; Inofuente-Ramos K.; Rosas-Paredes K.; Sulla-Torres J. · Proceedings of the 2025 IEEE 5th International Conference on Advanced Learning Technologies on Education and Research, ICALTER 2025 (2025) · DOI: 10.1109/icalter69698.2025.11355094

Resumen

A real-time phishing detection system is introduced that integrates an IDS with a character-level CharCNN classifier to analyze URL strings without relying on blacklists or page content. The development was structured with CRISP-DM and Scrum (four sprints: design, implementation, integration and evaluation) using a previously normalized and clean public dataset. Two configurations were trained (M1: lower/256; M2: both/512) and the capture-classify-report pipeline was validated in the laboratory. The results show improvement over the rules-only approach and stable operation in real time, with validation metrics close to 0.997 for M1, confirming technical feasibility and deployment potential in infrastructures with limited resources.

Instituciones: Universidad Católica de Santa María