Klasifikasi URL Berbahaya Menggunakan Algoritma Random Forest Berbasis Fitur Struktural

Abstract

Phishing attacks remain a critical threat in the digital era, often exploiting deceptive URLs to trick users into divulging sensitive personal information. To address this issue, this study proposes a machine learning-based detection system using the Random Forest algorithm to identify phishing URLs based on structural features. The main objective of this research is to build an efficient and lightweight model that can detect phishing attempts in real-time without relying on third-party databases or content-based analysis. From the dataset used, 10 structural features were selected based on relevance and efficiency, such as the presence of IP addresses, use of HTTPS, domain age, and URL length. The model was trained and tested on a labeled dataset and evaluated using accuracy, confusion matrix, and classification report. The Random Forest model achieved a testing accuracy of 92.72%, with strong precision and recall values for both phishing and legitimate classes. The results indicate that the proposed approach is effective in distinguishing malicious URLs using only structural characteristics, making it a practical solution for enhancing cybersecurity at the URL level.