Unofficial implementation of the network watermarking technique, "FINN". This code is not endorsed by the original authors and comes with aboslutely no guarentees or warranty.

@inproceedings{rezaei2021finn, title={FINN: Fingerprinting network flows using neural networks}, author={Rezaei, Fatemeh and Houmansadr, Amir}, booktitle={Proceedings of the 37th Annual Computer Security Applications Conference}, year={2021} }

Our implementation uses Python and Pytorch. Wireshark and pyshark are not required for the demo.

sudo apt install wireshark

conda create -n TrafficAnal python=3.12 pytorch=2.3.0 torchvision=0.18.0 torchaudio=2.3.0 torchtext=0.18.0 pytorch-cuda=12.1 scikit-learn scipy pandas scikit-learn matplotlib -c pytorch -c nvidia -c conda-forge

conda activate TrafficAnal

pip install pyshark

Downloading the CAIDA dataset is not required for the demo.

To get access to the CAIDA dataset, follow the instructions from CAIDA. After creating a USERNAME and PASSWORD, you can download the data using wget.

wget --directory-prefix=/PATH/TO/CAIDA --user=USERNAME --password=PASSWORD --recursive --level=16 --no-parent --no-clobber "https://data.caida.org/datasets/passive-2016/"
wget --directory-prefix=/PATH/TO/CAIDA --user=USERNAME --password=PASSWORD --recursive --level=16 --no-parent --no-clobber "https://data.caida.org/datasets/passive-2018/"

Optionally, get rid of all the uneeded files

find /PATH/TO/CAIDA \( -name "*.html*" -o -name "*.md5" -o -name "*.pdf" \) -type f -exec rm {} +

Processing the CAIDA dataset is not required for the demo.

After downloading the CAIDA datasets, we need to extract interpacket delay sequences from them. This is quite a lengthy process. On my server (40 cores), I found no benefit from using more than 8 python processes. It still took about a week to process all the data.

python src/caida.py --year="passive-2016" --source="equinix-chicago" --output="./data" --num_workers=8
python src/caida.py --year="passive-2018" --source="equinix-nyc" --output="./data" --num_workers=8

To produce a set of bash files containing the configuration for each of the experiments:

python ./run/reproduce.py --device=0

Add the --demo flag to run the demo only.

You can then run these individually using

CUDA_VISIBLE_DEVICES=0 bash ./run/{jobname}.sh

Or all at once using

./run/run.sh