PROCTOR: A Robust URL Protection System Against Fraudulent, Phishing, and Scam Activities

Abstract

Changes in internet usage patterns and behavior that have become increasingly massive since the COVID-19 pandemic have made hackers have various cybercrime ways to trick their victims. Some of the methods that are still used by hackers are fraud by utilizing user data with fake websites (phishing) that resemble the original website. The appearance and URL of the website that deceives the target or potential victim is a scam trick to gain the trust of the target. Therefore, we decided to research by building a URL detection system with the characteristics of fraud, phishing, and scam website-based using machine learning. Because this system is preventive in the form of protection, a user-friendly name was created, namely Protective URL Detector (PROCTOR). PROCTOR uses 52 standard features of website security protocols and is trained to leverage fraud, phishing, and scam data in Indonesia with random forest (RF) machine learning models. After training, the model is tested and evaluated with new data using the confusion matrix classification evaluation method. The most optimal model is achieved by the RF model with a training accuracy of 99.91