Mitigating Cybercrime Through Automated Darknet Traffic Detection Using CNNs

The rise in cybercrime activities on the darknet highlights the critical need for advanced techniques to detect and classify darknet traffic. The Tor network serves as a gateway to the darknet and dark web, enabling illicit activities such as unauthorized data exchanges, financial fraud, and cyberattacks. This paper proposes a novel approach to mitigating cybercrime through automated darknet traffic detection using Convolutional Neural Networks (CNNs). Our method employs automated encrypted payload extraction in hexadecimal format as distinguishing features, enabling the CNN model to capture unique encryption patterns inherent to Tor’s onion-routing protocol. The proposed model achieves 99.66% accuracy in identifying Tor traffic, demonstrating its effectiveness in darknet traffic classification. Additionally, a Random Forest-based feature importance analysis enhances computational efficiency by reducing training and inference times. The findings of this study contribute to cybersecurity by introducing a real-time, automated solution for mitigating cybercrime through darknet traffic detection.