A graphical deep learning approach to rf fingerprinting in the time-frequency domain

Bo Li*, Ediz Cetin

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review


The emergence of the internet of things (IoT) as a global infrastructure of interconnected network of heterogeneous wireless devices and sensors is opening new opportunities in myriad of applications. This growing pervasiveness of the IoT devices, however, is leading to growing concerns regarding security and privacy. Radio Frequency (RF) fingerprinting techniques operating at the physical layer can be used to provide an additional layer of protection to ensure trustworthy communications between devices to address these concerns. We present a graphical deep learning approach in the time-frequency (TF) domain based on short-time Fourier transform (STFT) where the intensity information of the STFT is used to generate two-dimensional image inputs for training and testing of the deep learning models for RF fingerprinting and identification. The performance of the proposed approach is evaluated and compared with the baseline approach operating in the waveform domain, using the same neural network architecture based on over-the-air captured datasets from 12 Zigbee devices. The experimental results show that the proposed approach outperforms both baseline approach, achieving near 100% identification accuracy based on data captured in a makeshift RF chamber, and accuracy of 98% on dataset which was captured under real-life conditions, demonstrating the robustness of the proposed approach. Further, the impact of the STFT-parameter selection on the identification performance of the proposed approach is also evaluated.
Original languageEnglish
Pages (from-to)1
Number of pages1
JournalIEEE Sensors Journal
Early online date5 Apr 2024
Publication statusE-pub ahead of print - 5 Apr 2024


  • deep learning
  • Internet of Things
  • RF fingerprinting
  • software defined radio
  • wireless device identification


Dive into the research topics of 'A graphical deep learning approach to rf fingerprinting in the time-frequency domain'. Together they form a unique fingerprint.

Cite this