AI-Based Phishing Detection Using Machine Learning

Abstract

Phishing attacks are one of the most common and continuously evolving cybersecurity threats, in which attackers use deceptive websites and URLs to trick users into revealing sensitive information such as usernames, passwords, financial details, and personal data. Traditional phishing detection techniques, including blacklists and rule-based approaches, may have difficulty identifying newly generated or previously unseen phishing URLs. Machine Learning (ML) provides an intelligent approach for automatically distinguishing legitimate URLs from potentially malicious ones based on their characteristics. This research proposes an AI-based phishing detection system that extracts relevant lexical and structural features from URLs and applies supervised machine learning algorithms to classify them as legitimate or phishing. The proposed approach can investigate and compare algorithms such as Logistic Regression, Decision Tree, Random Forest, Support Vector Machine, and Gradient Boosting based on evaluation measures including accuracy, precision, recall, F1-score, and ROC-AUC. Feature selection and preprocessing techniques will be applied to improve model performance and reduce unnecessary features. The study will use publicly available phishing and legitimate URL datasets for model training and evaluation. The objective is to identify an effective and computationally efficient machine learning model that can support early detection of phishing URLs. The proposed system has potential applications in web browsers, cybersecurity monitoring systems, email security, and educational security tools. The research also considers challenges such as dataset imbalance, changing phishing strategies, zero-day attacks, and model generalization, providing directions for developing more adaptive and reliable phishing detection systems.