Detecting URL Phishing Using BERT and DistilBERT Classifiers

In this paper, we investigate the use of transformer models, especially BERT and DistilBERT, for phishing URL detection. Such malicious URLs are a crucial part of phishing attacks, which trick users into disclosing personal information by sending them messages that seem to be from a trusted source. We suggest a novel method for URL phishing detection that uses developments in transformer-based natural language processing models. The primary goal of this study is to investigate the transformer-based models BERT and DistilBERT for the problem of URL phishing detection. The models’ excellent effectiveness in detecting phishing attempts highlights their potential for use in cybersecurity. The DistilBERT model fared better than its competitors, exhibiting exceptional accuracy, precision, recall, and F1-score metrics. The success of using cutting-edge NLP methods, particularly transformer models, to improve phishing detection systems.