DeepCRF_TIFS

This repository provides the implementation of DeepCRF. Read the paper for more details: DeepCRF: Deep Learning-Enhanced CSI-Based RF Fingerprinting for Channel-Resilient WiFi Device Identification

Setup

Requirements

This repository is built on Python 3.9.18 and Pytorch 2.1.2.
Other packages are listed in requirements.txt.

pip3 install -r requirement.txt

Datasets:

Dataset could be downloaded from this file: CSI Dataset.
After downloading, please move it to the data/ folder with the following directory structure:

$data/
  CSI/
  MAC/
  channel_BCDF.mat
  syn_testing_B_L.mat
  ...
  syn_testing_F.mat

Please note that the suffixes of the files of CSI data corresponding to P1-P9 in the paper are ['d3','d10','d12','d16','p1','p2','p3','outdoor',' mobilenlos'].

Training

# DeepCRF with contrastive loss
python3 main.py  Model='deepcrf-con' loss='contrastive'

# DeepCRF with ce loss
python3 main.py  Model='deepcrf' loss='cross'

# Baselines
python3 main.py  Model='ss' loss='cross' train_channel_num_per_channeltype=0 val_channel_num_per_channeltype=0
python3 main.py  Model='self-acc' loss='cross'
python3 main.py  Model='att_network' loss='cross'

Evaluation

# DeepCRF, evaluation with practical data
python3 evaluation.py  Model='deepcrf' test_with_practical_data=1 test_positions=['d3','d10','d12','d16','p1','p2','p3','outdoor','mobilenlos']

# DeepCRF, evaluation with synthetic data
python3 evaluation.py  Model='deepcrf' test_with_practical_data=0 channel_type=0 snr=40

Citation

@ARTICLE{kong2024deepcrf,
  author={Kong, Ruiqi and Chen, He},
  journal={IEEE Transactions on Information Forensics and Security}, 
  title={DeepCRF: Deep Learning-Enhanced CSI-Based RF Fingerprinting for Channel-Resilient WiFi Device Identification}, 
  year={2024},
  volume={},
  number={},
  pages={1-1},
  doi={10.1109/TIFS.2024.3515796}}