URL-Based Analysis for Detection of Fraudulent Websites Using Machine Learning

Abstract In this research, I developed a machine learning model to detect fraudulent websites using URL analysis. The dataset used in this study contained both legitimate and malicious URLs, which were collected from trusted public sources. I performed preprocessing to clean the data and carried out feature extraction, focusing on lexical and structural attributes such as URL length, number of digits, presence of HTTPS, subdomains, and special characters. These features helped the models in learning clear patterns that separate legitimate websites from fraudulent ones. I applied four machine learning algorithms: Decision Tree, Random Forest, Logistic Regression, and Naive Bayes. After training and evaluation, I found that Random Forest performed the best with an accuracy of 96.20%. The results showed that Random Forest was the most effective model for detecting fraudulent URLs, offering both accuracy and reliability. This research highlights the importance of machine learning and feature-based URL analysis in improving online security and protecting users from cyberfraud.