Malicious websites are a rich platform for many online attacks. These websites threaten the privacy and security of users, businesses, governments, and organizations every day. Malicious websites usually attempt to steal information or serve malware. Today's most popular type of malicious websites is phishing, which lure users into giving away sensitive information.

To protect users from such websites, detection capabilities need to be developed. Current detection methods such as signature based heuristic approaches are limited to known attack patterns and fail to scale with the rapid evolution of malicious websites on the web. This study collects a custom dataset of verified malicious phishing sites and uses machine learning classification techniques to detect malicious URLs and pages. Results show high recall and precision confirming that using machine learning techniques for detecting malicious websites can further improve the protection of users on the web.