Paper Accepted in Computer Networks: SMoRFFI Dataset and RF Fingerprinting Framework

We are pleased to announce that our paper, “SMoRFFI: A Large-Scale Same-Model 2.4 GHz Wi-Fi Dataset and Reproducible Framework for RF Fingerprinting”, has been officially published in Computer Networks (2026).

Overview

Radio frequency (RF) fingerprinting has emerged as a promising approach for device identification by exploiting hardware imperfections at the physical layer. However, identifying same-model devices remains an open and challenging problem, as such devices exhibit highly similar RF characteristics.

To address this challenge, we present SMoRFFI, a large-scale dataset and a fully reproducible experimental framework specifically designed for same-model RF fingerprinting.

What We Provide

  • A large-scale same-model dataset:
    • 123 commercial IEEE 802.11g devices
    • 35.42 million raw I/Q samples from Wi-Fi preambles
    • 1.85 million extracted RF features
  • A rich feature representation:
    • Frequency-related features (CFO, coarse/fine CFO)
    • Constellation-based features (phase error, magnitude error)
    • Hardware-related features (I/Q imbalance, fractal dimension
  • A fully reproducible pipeline:
    • Data collection (USRP + GNU Radio)
    • Feature extraction
    • Benchmark evaluation

Key Insights

  • Same-model RF fingerprinting is significantly more difficult than heterogeneous-device scenarios due to minimal hardware variation
  • Frequency-related features (e.g., CFO) are the most discriminative for device identification
  • Even with a lightweight model, our baseline achieves 88.6% accuracy, highlighting both:
    • the difficulty of the dataset, and
    • the potential for more advanced models

Why This Work Matters

Unlike existing datasets, SMoRFFI:

  • Focuses on the worst-case scenario (same-model devices)
  • Provides both raw I/Q data and pre-extracted RF features
  • Ensures full reproducibility with open-source tools

This makes it a strong benchmark for:

  • RF-based authentication
  • Open-set / unknown device detection
  • Machine learning and deep learning on RF signals

Resources

Citation

Guo, Z., Jia, Z., Zhu, J., Huang, W., & Chen, Y. (2026).
SMoRFFI: A large-scale same-model 2.4 GHz Wi-Fi dataset and reproducible framework for RF fingerprinting.
Computer Networks, 112309.


We hope this work provides a solid foundation for future research in RF fingerprinting and secure wireless device identification.