User-Centric Web Application for Phishing URL Detection by Machine Learning Model

Phishing attacks threaten individuals and organizations globally. While existing phishing detection tools predominantly focus on technical aspects, there exists a notable void in addressing the informational needs of users beyond numerical data. Additionally, there is a lack of studies focusing on users and few have explored feature selection techniques in machine learning-based phishing detection. To bridge this gap, we have developed a user-centric phishing URL detection tool powered by machine learning models. The objective is twofold: to effectively categorize phishing links and to provide users with contextual understanding and educational resources to enhance their online safety. This paper offers unique contributions by a comprehensive, user-focused approach to phishing detection combined with ML techniques. For the ML aspects, three feature selection approaches—Recursive Feature Elimination with Cross-Validation (RFECV), Particle Swarm Optimization (PSO), and Random Forest selection—were explored using three different machine learning algorithms: Random Forest, LightGBM, and SVC. The study found that the combination of LightGBM with RFECV provided the highest performance, achieving an accuracy of 95.07% after hyperparameter tuning. By focusing on user-centric design, this research not only enhances phishing detection capabilities but also empowers users with the knowledge to make informed decisions, thereby improving overall online safety.