The increasing prevalence of cyber threats has made the identification and mitigation of harmful URLs an essential cybersecurity challenge. This research presents a machine learning-based framework that categorizes URLs as either safe or harmful. This study dives into the effectiveness of different classification techniques, including Naive Bayes, Logistic Regression, SGD Decision Tree Classifier, and Support Vector Machines (SVM). The models were trained on datasets featuring lexical, host-based, and behavioral indicators. Results indicate that combining these extracted attributes with machine learning enhances detection performance. This comparative analysis provides key insights into selecting the most effective algorithm for cybersecurity applications.