POWDER RF Fingerprinting (RFP)

Device-level RF fingerprinting captures from the POWDER PAWR platform.

Official links

• Raw data: https://www.genesys-lab.org/powder
• Paper: Trust in 5G Open RANs through Machine Learning (Globecom 2020)

Tasks supported

• RF fingerprinting (Mean Per-Class Accuracy)

Class labels

• bes
• browning
• honors
• meb

Split

• 80/20 random split

Preprocessing

1. Match .bin/.json pairs by stem and filter for complex64 IQ recordings.
2. Assign transmitter labels from JSON (annotations.transmitter.core:location).
3. Slice each recording into fixed-length chunks (default 512) with optional hop.
4. Compute global I/Q mean and std across all chunks (first pass).
5. Normalize each chunk, then write IQ, label, and metadata (sample rate, center freq, offsets) to HDF5.

Script: preprocess_rfp.py

python preprocessing/preprocess_rfp.py \
  --data-path <POWDER_DIR> \
  --output data/rfp.h5 \
  --chunk-len 512

Metric

• Mean Per-Class Accuracy on the test split.

Citation

@inproceedings{reusmuns2019trust,
  title={Trust in 5G Open RANs through Machine Learning: RF Fingerprinting on the POWDER PAWR Platform},
  author={Reus-Muns, Guillem and Jaisinghani, Dhertya and Sankhe, Kunal and Chowdhury, Kaushik},
  booktitle={IEEE Globecom 2020-IEEE Global Communications Conference},
  year={2020},
  organization={IEEE}
}