Malicious URL Detection: A Comparative Study

Malicious uniform resource locator (URL), i.e., Malicious websites are one of the most common cybersecurity threats. They host gratuitous content (spam, malware, inappropriate ads, spoofing, etc.) and tempt unwary users to become victims of scams (financial loss, private information disclosure, malware installation, extortion, fake shopping site, unexpected prize etc.) and cause loss of billions of rupees each year. The visit to these sites can be driven by email, advertisements, web search or links from other websites. In each case, the user must click on the malicious URL. The rising cases of phishing, spamming and malware has generated an urgent need for a reliable solution which can classify and identify the malicious URLs. Traditional classification techniques like blacklisting, regular expression, and signature matching approach are challenged because of huge data volume, patterns and technology changing over time, along with complicated relationship among features. In this paper, we address the detection of malicious URLs as a binary classification problem and evaluate the performance of several well-known machine learning classifiers. We adopted a public dataset from Kaggle comprising of 450000 URLs to train the model. The best classifier was used to detect malicious URLs from openphish website. It was found to give better results.