Proactive Phishing Defense: A URL Classification System Using Machine Learning

Phishing attacks are the most common cyberattacks nowadays. Phishing attacks rely on social engineering concepts to trick victims into reaching the goals of malicious attackers. In addition, phishing attacks are the largest vector for various cyberattacks. However, URLs are a fulcrum for phishing attacks. The difficulty distinguishing between legitimate and phishing URLs is the reason for the increased success rates of these attacks. An integrated framework is proposed in this study to detect phishing attacks based on classifying URLs into phishing or legitimate URLs through machine learning models such as decision tree (DT) and random forest (RF), which have high power and prediction accuracy in binary classification tasks. The RF model, using the cross validation (CV) technique, achieved an accuracy score of $\mathbf{9 8. 2}$. This methodology is embedded in a web application with a graphical user interface to provide ease of handling and show alerts in real time and visually. This contributes to providing the field of cybersecurity with a highly accurate verification system to reduce users falling victim to these dangerous attacks.