Capture and logging of Tor cells.

The Tor source code is modified to directly emit cell sequences to a logger using UDP packets for use in website fingerprinting experiments. Tor is run in docker and restarted for each website capture to completely reset any existing circuits.

pip3 install -r requirements.txt

wget https://github.com/mozilla/geckodriver/releases/download/v0.31.0/geckodriver-v0.31.0-linux64.tar.gz
tar xvzf geckodriver-v0.31.0-linux64.tar.gz
chmod +x geckodriver
mv geckodriver /usr/local/bin

follow installation directions according to https://docs.docker.com/engine/install/

The modified instance of Tor is used to emit cell sequences of a website trace to the cell logger. Tor is run in docker for this experiment.

docker build -t tor .

The cell logger captures and stores the cells emitted from Tor for each trace in a .cell file delimited by a tab.

make

The controller will look for any extensions implemented using Mozilla Browser Extensions in the ./visit/addon/ directory. An example is here which implements half duplex communication for the Walkie-Talkie and Tail Time website fingerprinting defenses. To add any extensions to the capture, place their .xpi file in the ./visit/addon/ directory.

The websites captured consists of those listed in the websites.txt file. A sample is provided. In the original work, the top 100 websites from the Alexa Top Sites (this link will start a download) are used after removing duplicate and localized domains.

The controller will do 100 captures in batches of 25 for each website in the websites.txt file. It will do this for each extension in the ./visit/addon/ directory as well as once for normal traffic without any extensions.

bash controller.sh

The traces need to be sorted since they are collected using UDP packets.

python3 sort_traces.py

J. Liang, C. Yu, K. Suh and H. Han, "Tail Time Defense Against Website Fingerprinting Attacks," in IEEE Access, vol. 10, pp. 18516-18525, 2022, doi: 10.1109/ACCESS.2022.3146236.