/RFonEdge

Primary LanguagePython

RFonEdge

The code in this repository implements the algorithms and experiments in the following paper:

T. Jian, Y. Gong, Z. Zhan, R. Shi, N. Soltani, Z. Wang, J. Dy, K. Chowdhury, Y. Wang, S. Ioannidis, “Radio Frequency Fingerprinting on the Edge”, IEEE Transactions on Mobile Computing, 2021.

Prerequisites

pytorch-1.6.0
torchvision-0.7.0
numpy-1.16.1
scipy-1.3.1
tqdm-4.33.0
yaml-0.1.7

Experiments

We implement progressive model pruning algorithm that progressively prune the pre-trained model that satisfies the pre-defined sparsity constraint sets for filter or column pruning.

We evaluate our algorithm on five benchmark datasets, including three WiFi datasets (WiFi-50, WiFi-Eq-50, WiFi-Eq-500), one ADS-B dataset (ADS-B-50), and one mixture dataset (Mixture-50) containing both WiFi and ADS-B transmissions.

We run all algorithms via main.py, and provide several useful tools to define/check sparsity settings as follows:

  • testers.py for quick checking of the sparsity.

  • flops.py for quick checking of model FLOPS.

  • profile/config*.yaml template the configuration files. Each represents a resulting pruning ratio.

  • run.sh templates an example script for running the code.

Citing This Paper

Please cite the following paper if you intend to use this code for your research.

T. Jian, Y. Gong, Z. Zhan, R. Shi, N. Soltani, Z. Wang, J. Dy, K. Chowdhury, Y. Wang, S. Ioannidis, “Radio Frequency Fingerprinting on the Edge”, IEEE Transactions on Mobile Computing, 2021.

Acknowledgements

Our work is supported by National Science Foundation under grants CCF-1937500 and CNS-1923789.