Protecting electronic systems from malicious attacks is a critical con cern in cybersecurity. However, human error remains a significant vul nerability, often leading to security breaches through the inadvertent clicking of malicious URLs. These URLs are disseminated through various channels, including social media platforms and email, and can introduce viruses, malware, and other harmful programs, jeopardizing user data and device functionality. Consequently, it is imperative for organizations to develop effective methods for detecting and prevent ing the proliferation of such malicious URLs. This study aims to enhance the accuracy of malicious URL detec tion by identifying essential attributes that contribute to accurate classification. We conduct a comparative analysis of different ma chine learning algorithms to determine the most effective approach for URL detection. Additionally, we propose a novel input strategy by incorporating extracted features from the URL structure and con tent to further enhance the classification process. Furthermore, we explore feature selection techniques to identify the most informative attributes for accurate classification. Preliminary results demonstrate promising outcomes, with the Ran dom Forest algorithm achieving the highest accuracy among the tested machine-learning algorithms. Moreover, our new approach, leveraging additional input, showcases improved accuracy while reducing train ing time. These findings underscore the potential of our method to enhance the detection of malicious URLs and mitigate the risks asso ciated with human error and cyber threats.