Transformer Based Method for Attack Detection in Malicious URLs

Abstract

Transformer Based Method for Attack Detection in Malicious URLs Abstract: The detection of malicious URLs is paramount in cybersecurity due to their pivotal role in enabling various cyber threats, including malware distribution, phishing scams, and website defacement. This paper introduces an innovative approach leveraging transformer-based model the DistilBERT architecture, to tackle this critical challenge. Our method is built upon a meticulously curated dataset, encompassing URLs that are labeled into distinct categories such as malware, phishing, defacement, and benign URLs. Through extensive experimentation and validation, our approach demonstrates a remarkable accuracy rate of 99.35\%, underscoring its effectiveness. The results of this study provide a reliable and effective method for the identification of malicious URLs, which significantly advances the subject of cybersecurity. This in turn enhances online defence mechanisms against a broad spectrum of cyber-attacks. Our work not only highlights the potential of transformer-based model in cybersecurity applications but also provides a practical framework for deploying the advanced technique to safeguard against malicious online activities.