Website Fingerprinting Through the Cache Occupancy Channel and its Real World Practicality

Abstract

Website fingerprinting attacks, which use statistical analysis on network traffic to compromise user privacy, have been shown to be effective even if the traffic is sent over anonymity-preserving networks such as Tor. The classical attack model used to evaluate website fingerprinting attacks assumes an on-path adversary, who can observe all traffic traveling between the user's computer and the secure network.In this work we investigate these attacks under a different attack model, in which the adversary is capable of sending a small amount of malicious JavaScript code to the target user's computer. The malicious code mounts a cache sidechannel attack, which exploits the effects of contention on the CPU's cache, to identify other websites being browsed. The effectiveness of this attack scenario has never been systematically analyzed, especially in the open-world model which assumes that the user is visiting a mix of both sensitive and non-sensitive sites.We show that cache website fingerprinting attacks in JavaScript are highly feasible. Specifically, we use machine learning techniques to classify traces of cache activity. Unlike prior works, which try to identify cache conflicts, our work measures the overall occupancy of the lastlevel cache. We show that our approach achieves high classification accuracy in both the open-world and the closedworld models. We further show that our attack is more resistant than network-based fingerprinting to the effects of response caching, and that our techniques are resilient both to network-based defenses and to side-channel countermeasures introduced to modern browsers as a response to the Spectre attack. To protect against cache-based website fingerprinting, new defense mechanisms must be introduced to privacy-sensitive browsers and websites. We investigate one such mechanism, and show that generating artificial cache activity reduces the effectiveness of the attack and completely eliminates it when used in the Tor Browser.

Undermining User Privacy on Mobile Devices Using AI

“…One reason for these differences is that we incorporated the results by Prouff et al [41] in addition to conducting preliminary experiments. The classification rates for Google Chrome in [44] are comparable to the ones we achieve. In summary, both Shusterman et al and this work propose inference attacks that share a common goal, but differ regarding approach and attack environment.…”

“…Compared to the LLC-based attack in [39], our success rates are higher, even though we classify significantly more websites. Compared to the results by Shusterman et al [44], we achieve similar classification rates. At the same time, we relax the attacker model by not only compensating imprecise timing sources but also random cache replacement policies.…”

“…In [44], the authors present a similar inference attack than the one proposed in this work. The LLC profiling is implemented with so-called one-dimensional memorygrams.…”

“…Shusterman et al [44] extend this work by inferring websites from JavaScript with simple last-level cache profiles that are classified by Convolutional Neural Networks and Long Short-Term Memory. As this is concurrent work to ours 3 , we provide a closer comparison later in the section.…”

“…We trained our model with 2 Convolution layers, which yield a lower validation loss compared to 3 layers. The kernel size is kept the same in LLC 78.5 40 [53] iOS APIs 89.0 120 [46] Java-based Android API 85.6 20 [9] Interrupt Handling Mechanism 87.0 100 [45] ProcFS Leaks 96.0 100 each Convolution layer, while in [44] it is varied with each layer. Finally, we use a Pooling size of 2, while Shusterman et al use 4.…”