GAN-based Tor obfuscated traffic identification on imbalanced datasets

This research paper examines how GANs (generative adversarial networks, machine learning models that learn to generate realistic data) can identify Tor traffic (encrypted network connections that hide user identity) even when that traffic has been obfuscated (disguised to look like normal internet activity). The study addresses the challenge of working with imbalanced datasets (training data where one category vastly outnumbers others), which is common in real-world network security scenarios.