Advanced Phishing Detection Using Hybrid Model

Abstract

Phishing attacks are growing in number at a rapid pace and represent a serious threat.New malware types are a serious Cybersecurity problem, especially as the new types appear as sophisticated and AI generated phishing URLs.Traditional Based methods use anti-phishing tools to identify phishing sites.Anti-phishing tools utilize phishing detection techniques such as blacklist-based and rule-based methods to detect newly created, based systems,zero-day attacks. This paper introduces a Hybrid Transformer-Based URL phishing detection system, which is a combination of deep and shallow learning systems.To enhance phishing detection, the proposed framework aligns lexical URL feature extraction.The transformer model was used to create embeddings.The random forest classifier is used to fuse and classify these features.To detect phishing URL from valid URLs, the URLs are effectively passed to the classifier.The system is trained and evaluated using PhiUSIIL phishing data set. Experimental results show that the proposed model is effective in outperforming other models.Performance with high accuracy, precision, recall and F1 score was obtained compared to conventional machine learning and ensemble-based approaches. The proposed system is a good solution to detect AI-created phishing URLs, improved Flexibility to changing phishing tactics. The framework can be used in real-time for deployment in web browsers enterprise Cybersecurity applications