HOT is a made-up short for the Swedish project name ‘Hoppet till Tor’, roughly translatable to ‘The hope/jump to Tor’. The project focused on the first connection from a user’s computer into the Tor network. The goal of the project was to experimentally evaluate and customize one or more pluggable transports (PTs) for Tor to protect against website fingerprinting (WF) attacks. HOT was funded by the Swedish Internet Fund and ran throughout 2016. The principal investigator, Tobias Pulls, was also partially funded by the excellence environment at Karlstad university.

During the first four months of the project, we focused on better understanding the setting of WF attacks, by:

• looking at the WF setting,
• the basic measurements,
• the more advanced measurements, and
• the kNN attack in detail.

The project then took a de-tour, with a focused contribution to work on understanding the effect of DNS on Tor’s anonymity together with researchers at Princeton University and KTH Royal Institute of Technology. During this time, since our work was to a large extent on new attacks on Tor, we decided to not be open with our research until we knew the implications of our findings. Beyond our project page—with open data, code, and an academic paper—our work was featured on blogs and news articles world-wide (see, e.g., ZDNet, The Register, and IDG). Finally, our paper has also been accepted at the NDSS Symposium 2017. The attacks we present show how an attacker could use observed DNS traffic made from the Tor network to enhance WF attacks to be more precise. As part of the work on DNS and Tor, we developed a large number of tools and collected two large datasets:

• the DefecTor DNS dataset, and
• the DefecTor WF dataset.

For the last four months, we were back on track and decided to focus on basket2: a candidate for the next generation of pluggable transport in Tor. In a series of posts, we evaluated the current padding methods as implemented in basket ([1,2,3]) using tools we developed that were customized for evaluating pluggable transports (1, 2, 3). Since our project proposal was written, a clear candidate for Tor as a WF defense emerged, in the form of WTF-PAD. Lacking a public implementation, we decided to implement a basic version of WTF-PAD into basket2—which we named Adaptive Padding Early (APE)—and evaluated it using our tools. Our hope is that our work will serve as a step towards a widely deployed—by default—WF defense in Tor.