Malicious URL Detection and Categorization Using Machine Learning Techniques

The proliferation of online businesses has also given rise to a growing number of malicious activities, with compromised Uniform Resource Locators (URLs) being a significant threat to individuals and organizations alike. Malicious URLs serve as vectors for various cyberattacks, including phishing, malware distribution, and drive-by-downloads, resulting in financial and data losses. Traditional detection methods, such as blacklist-based systems, have limitations in coping with the dynamic nature of these threats. Consequently, machine learning models have emerged as a viable solution to improve the detection and classification of malicious URLs. This study reviews various machine learning approaches, including Random Forest, Logistic Regression, Support Vector Machines, and neural networks, applied to detect malicious URLs. The literature reveals that Random Forest consistently performs well, achieving accuracies above 99% in several cases. This paper underscores the importance of feature selection, dataset balance, and classifier performance. Additionally, it highlights the need for adaptive models that can keep pace with evolving cyberattacks.