Device RF Fingerprint

RF fingerprinting for device identification: classify which device transmitted a signal from raw I/Q or preprocessed features. Positioned for product-oriented use (e.g. IoT security, access control).

Overview

Each transmitter has slight hardware imperfections that leave a unique "fingerprint" in the radio signal. This project trains a classifier on I/Q traces (or hand-crafted features) to identify the device. A Gradio demo lets users try the model interactively.

Data

Use a public RF fingerprint dataset (e.g. WiFi/USRP-based) or synthetic data for demonstration.
Format: (samples, 2, seq_len) for I/Q or (samples, features) for feature-based input.

Usage

Training:

pip install -r requirements.txt
python train.py

Demo:

python app.py

Then open the Gradio URL to upload or paste signal data (or use synthetic) and get a device ID prediction.

Model

CNN on I/Q windows. Implemented in model.py; training in train.py; Gradio demo in app.py. Checkpoints go to ./checkpoints/rf_fingerprint.pt.

Limitations / future work

Performance depends on channel and hardware stability; real deployments may need periodic retraining.
Could add confidence scores and calibration for security-critical use.

Author

Alireza Aminzadeh

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support