A novel application classification attack against Tor

摘要: Tor is a famous anonymous communication system for preserving users' online privacy. It supports TCP applications and packs upper-layer application data into encrypted equal-sized cells with onion routing to hide private information of users. However, we note that the current design cannot conceal certain behaviors. For example, P2P usually upload download files simultaneously, this behavioral feature also kept in traffic. Motivated by observation, investigate new attack against Tor, classification attack, which can recognize types from An attacker first carefully selects some flow features such as burst volumes directions represent behaviors takes advantage efficient machine-learning algorithm e.g., Profile Hidden Markov Model model different applications. Then he or she use these established models classify target's traffic infer its type. We have implemented on using parallel computing, our experiments validate feasibility effectiveness attack. argue disclosure type serious threat anonymity because it be used reduce set facilitate other attacks. present guidelines defend Copyright © 2015 John Wiley & Sons, Ltd.