Survey paper on Phishing Website Detection using Supervised Machine Learning Technique

Abstract

Phishing websites have become a major cybersecurity threat, targeting users by impersonating legitimate websites to steal sensitive information such as usernames, passwords, banking credentials, and personal data. Traditional blacklist- and rule-based detection methods often struggle to identify newly generated and rapidly changing phishing websites. In recent years, supervised machine learning (ML) techniques have emerged as effective approaches for detecting phishing websites by learning patterns from labeled datasets containing legitimate and malicious website characteristics. This survey paper presents a comprehensive review of supervised ML techniques used for phishing website detection, including Decision Tree, Random Forest, Support Vector Machine, K-Nearest Neighbors, Naïve Bayes, Logistic Regression, and ensemble learning methods. The survey examines commonly used phishing-related features, such as URL structure, domain information, hyperlink characteristics, webpage content, security indicators, and HTML-based attributes. Furthermore, the study compares existing approaches based on evaluation metrics including accuracy, precision, recall, F1-score, and ROC-AUC. The advantages and limitations of different supervised learning methods are discussed, with particular emphasis on feature selection, class imbalance, dataset quality, and generalization to previously unseen phishing websites. The survey also identifies current research challenges and potential future directions, including hybrid ML models, automated feature engineering, explainable artificial intelligence, and real-time phishing detection systems. This study provides a structured overview of supervised machine learning approaches and highlights promising directions for developing accurate, scalable, and robust phishing website detection systems.