Improving Phishing Website Detection with Machine Learning: Revealing Hidden Patterns for Better Accuracy

Abstract

Phishing attacks remain a significant threat to internet users globally, leading to substantial financial losses and compromising personal information. This research study investigates various machine learning models for detecting phishing websites, with a primary focus on achieving high accuracy. After an extensive analysis, the Random Forest Classifier emerged as the most suitable choice for this task. Our methodology leveraged machine learning techniques to uncover subtle patterns and relationships in the data, going beyond traditional URL and content-based restrictions. By incorporating diverse website features, including URL and derived attributes, Page source code-based features, HTML JavaScript-based features, and Domain-based features, we achieved impressive results. The proposed approach effectively classified the majority of websites, demonstrating the efficiency of machine learning in addressing the phishing website detection challenge with an accuracy of over 98%, recall exceeding 98%, and a false positive rate of less than 4%. This research offers valuable insights to the field of cyber security, providing internet users with improved protection against phishing attempts.