BERT-LogReg: Enhancing Phishing URL Detection With Transformr-Based Features

Phishing URLs are fraudulent websites created to trick visitors into downloading malware or divulging personal information. Identifying phishing URLs is essential to maintaining internet security and safety. However, blacklists and keyword matching are frequently used in classical detection techniques, which are readily manipulated by adversaries. This study suggests a unique method for phishing URL identification that makes use of transformers and machine learning. With our method, phishing and authentic websites may be reliably distinguished by automatically extracting and learning discriminative characteristics. We assess our method using an extensive dataset of authentic and phishing URLs from the real world and show that it can identify phishing URLs that have never been seen before. Our findings imply that deep learning may greatly increase the efficacy and accuracy of phishing URLs.