Intelligent Malicious URL Detection Using Extreme Multi-Label Text Classification for Phishing Prevention

Attackers employ phishing to take advantage of people's lack of awareness and knowledge about the internet. An attacker may, for instance, construct a phony website that appears to be an online store in order to obtain sensitive info such as customers' names, addresses, phone numbers, besides credit card particulars. An extension to protect internet users from phishing attempts is suggested in this research. Using deep learning models, this study specifically builds a smart warning feature for the proposed extension. By scanning for, alerting about, and blocking unsafe connections, the suggested browser extension safeguards users. In order to improve the detection accuracy and scalability, this study suggests a Malicious URL Phishing Attack Detection Model that utilizes the Extreme Multi-label Text Classification (XMTC) Algorithm. To accurately identify harmful links, the suggested method parses URLs for textual properties and then sorts them into various phishing hazard categories. Using score, metrics, the model is trained on a huge dataset that contains both authentic and phishing URLs. The results demonstrate that XMTC improves classification performance and easily handles highdimensional data, making it superior to traditional machine learning-based phishing detection methods. According to the research, the XMTC-based model is an intelligent phishing detection system that is scalable, strong, and can mitigate cyber risks in realtime online settings. The purpose of this study was to compare and contrast the effectiveness of several machine learning models by utilizing a malicious URL dataset with 651,191 data samples. The investigation's findings corroborate the suggested extension's excellent accuracy, which ranged from 96% to 97%.