Effectiveness of Website Fingerprinting in Real-World Settings

Determine whether Website Fingerprinting attacks against the Tor network can be effective in real-world settings beyond controlled laboratory conditions.

Background

Website Fingerprinting attacks infer which webpages a user visits by analyzing encrypted traffic, such as traffic carried over the Tor anonymity network. Although machine-learning-based attacks have achieved nearly 100% accuracy in laboratory conditions, those results may depend on restrictive experimental assumptions, including a single machine, a fixed browser, a stable collection period, and a limited set of monitored individuals or webpages.

The paper explicitly identifies the transferability of Website Fingerprinting effectiveness to real-world conditions as unresolved. Empirical measurements cited by the paper suggest that such attacks may succeed only under special conditions, leaving open whether they can operate effectively in broader and more realistic deployment scenarios.