Machine Learning-Based Detection of Phishing URLs with Advanced Hyperlink Analysis

Abstract

This research detects phishing URLs using machine learning algorithms that analyze hyperlinks present on a website. The method incorporates novel hyperlink-specific features that are divided into 30 categories to enhance its accuracy in identifying phishing attacks. The proposed approach is entirely client-side and can detect websites written in any language. The effectiveness of the proposed approach was evaluated by testing it on a dataset containing both phishing and non-phishing websites using various classification algorithms. The Random Forest classifier achieved over 97.4% accuracy in detecting phishing websites. The study suggests that feature selection and classifier choice are essential in detecting malicious URLs.

The research highlights the importance of detecting malicious websites to protect users against criminal activities. As web applications grow in importance, cybercriminals are becoming more sophisticated, and detecting malicious websites is crucial to protecting users. The proposed technique can be combined with different feature extraction models to test its usefulness in real-time scenarios for automatic detection of websites and web browser extensions.

The research shows that machine learning algorithms can effectively detect phishing URLs and that further improvements can be made by incorporating more features, expanding the dataset, and using different classification techniques.