Real-Time Detection of Malicious URLs Using Feature-Based Machine Learning Approaches

Authors: Hrushikesh Ghuge, Disha Ghuge, Anmol Rangneniwar, Parth Samudra, Dr. A.V. Markad

Country: India

Full-text Research PDF File:   View   |   Download

Abstract: Malicious URLs serve as primary vectors for cyber threats such as phishing, defacement, and malware attacks. Traditional blacklist-based detection methods fail to identify newly emerging threats, necessitating the use of machine learning techniques for improved detection accuracy. In this study, we propose a Random Forest-based classification model for malicious URL detection, utilizing lexical and structural features extracted from URLs. The dataset was balanced to ensure fair training across all classes, including benign, defacement, phishing, and malware URLs. The trained model achieved an accuracy of 88.32%, with high precision and recall for defacement detection. While the model demonstrates promising results, further improvements in feature engineering and dataset diversity could enhance detection performance against evolving threats.

Keywords: Malicious URL, Cybersecurity Threat, Phishing, Malware Propagation, Blacklist, Heuristics, Machine Learning, Deep Learning, Data Scarcity

Paper Id: 232401

Published On: 2025-04-21

Published In: Volume 13, Issue 2, March-April 2025

Share this