This paper presents a phishing URL detection system using a fine-tuned BERT-based multimodal approach. The model combines textual analysis of URLs with structural features such as URL length, number of digits, special characters, and HTTPS presence. The objective is to improve generalization and detection accuracy for phishing attacks. The system is evaluated on a sample dataset and demonstrates effective classification between phishing and legitimate URLs. This work highlights the application of deep learning in enhancing cybersecurity solutions.
Ingale et al. (Tue,) studied this question.