A Machine Learning-based Framework for Realtime Tor Traffic Detection

The Tor network enables anonymous communication by encrypting and routing traffic through relays, providing privacy for users in censored regions. However, Tor is also exploited for illegal activities, such as cyberattacks and illicit online marketplaces. Consequently, detecting Tor network-based traffic while preserving user privacy is a critical and promising approach to address this challenge. In this paper, we propose the Machine Learning-based Framework for Realtime Tor Traffic Detection (MLF-RTTD), which leverages statistical features extracted solely from packet headers, without analyzing packet payloads. Our experimental results indicate that MLF-RTTD achieves a high classification accuracy of 97%, clearly demonstrating its potential as a realtime, privacy-preserving solution to the Tor traffic detection problem.