k-fingerprinting: A Robust Scalable Website Fingerprinting Technique

Website fingerprinting enables an attacker to infer the source of a web page when a client is browsing through encrypted or anonymized network connections. We present a new website fingerprinting attack based on fingerprints extracted from random decision forests and evaluate performance on three separate data sets consisting of both standard web pages as well as Tor hidden services. Within the context of this attack we provide an analysis of the utility of previously proposed traffic features. Our attack, k-fingerprinting, performs better than current state-of-the-art attacks even against website fingerprinting defenses. We show that it is possible to launch a website fingerprinting attack in the face a large amount of noisy data. We further show that error rates vary widely between web resources, and thus some patterns of use will be predictably more vulnerable to attack than others.