Developing a Model to Detect Malicious URLs using Different Classification Algorithms

The number of social media users in Saudi Arabia is increasing by the day, and the rate of cyberattacks has increased accordingly. This can be partially due to the dearth of tools that can be used to check and ensure the integrity of URLs. This study introduced a model for classifying URLs into benign and malicious ones to help users avoid falling victim to phishing schemes using URLs contained in messages. By using five classification algorithms were compared: logistic regression (LR), support vector machine (SVM), random forest (RF), decision tree (DT) and gradient boosting (GB). The algorithms’ performances were based on their accuracy and confusion matrices. The SVM algorithm was found to be the most accurate (95.4%) and the best algorithm for verifying the validity of URLs.