A Critical Evaluation of Website Fingerprinting Attacks

Abstract

Recent studies on Website Fingerprinting (WF) claim to have found highly effective attacks on Tor. However, these studies make assumptions about user settings, adversary capabilities, and the nature of the Web that do not necessarily hold in practical scenarios. The following study critically evaluates these assumptions by conducting the attack where the assumptions do not hold. We show that certain variables, for example, user's browsing habits, differences in location and version of Tor Browser Bundle, that are usually omitted from the current WF model have a significant impact on the efficacy of the attack. We also empirically show how prior work succumbs to the base rate fallacy in the open-world scenario. We address this problem by augmenting our classification method with a verification step. We conclude that even though this approach reduces the number of false positives over 63%, it does not completely solve the problem, which remains an open issue for WF attacks.

Toward an Efficient Website Fingerprinting Defense

“…Even with the challenges of open-world and multi-tab browsing [11], some websites may exhibit especially unique traffic patterns and be prone to high-confidence attacks. Attacks may observe visits to the same site over multiple sessions and gain confidence in a result.…”

“…HTTPOS modifies HTTP headers and injects HTTP requests strategically [13], while Randomized Pipelining, a WF countermeasure currently implemented in the Tor Browser, randomizes the pipeline of HTTP requests. Both defenses have been shown to be ineffective in several evaluations [5,24,23,11].…”

“…This approach aims to optimally reduce the amount of padding needed to confound the attacker's classifier. These defenses, as well as traffic morphing techniques [25,12], have the shortcoming that require a database of webpage templates that needs to be frequently updated and would be costly to maintain [11].…”

“…We used a dataset that had been collected for a study about a realistic evaluation of WF attacks [11]. This dataset consists of 40 instances, collected in ten batches of four visits, for each homepage in top-100 Alexa sites [1].…”

Website Fingerprinting Defenses at the Application Layer

“…We therefore implement a simple client-side WF defense, dubbed Lightweight application-Layer Masquerading Add-on (LLaMA), that works at the application layer by adding extra delays to the HTTP requests. These delays alter the order of the requests in a similar way to randomized pipelining (RP) [23], a WF countermeasure implemented in the Tor browser that has been shown to fail in several evaluations [5,14,31]. Besides delaying HTTP requests, our defense sends redundant requests to the server.…”