Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Tracking UWB Devices Through Radio Frequency Fingerprinting Is Possible

Thibaud Ardoin, Niklas Pauli, Benedikt Groß, Mahsa Kholghi, Khan Reaz, Gerhard Wunder

TL;DR

This work investigates whether Radio Frequency Fingerprinting can be applied to Ultra-Wideband signals for device identification. By collecting a large, controlled dataset of 1.5 million+ measurements using 14 UWB boards and applying an open-set re-identification framework with CNN and Vision Transformer (ViT) models, the study demonstrates near-perfect accuracy in stable, fixed setups and meaningful generalization challenges when environments or devices vary. The authors introduce ArcFace loss within the ViT architecture to improve discriminability and show that multi-sample inputs and open-set evaluation can enhance robustness, while noting significant generalization gaps across days, locations, and unknown devices. The contributions include a pioneering UWB RFF study, a realistic dataset with controlled variation, and an enhanced DL pipeline with empirical results, highlighting both the security potential and privacy concerns of hardware-based fingerprinting in UWB systems.

Abstract

Ultra-wideband (UWB) is a state-of-the-art technology designed for applications requiring centimeter-level localization. Its widespread adoption by smartphone manufacturer naturally raises security and privacy concerns. Successfully implementing Radio Frequency Fingerprinting (RFF) to UWB could enable physical layer security, but might also allow undesired tracking of the devices. The scope of this paper is to explore the feasibility of applying RFF to UWB and investigates how well this technique generalizes across different environments. We collected a realistic dataset using off-the-shelf UWB devices with controlled variation in device positioning. Moreover, we developed an improved deep learning pipeline to extract the hardware signature from the signal data. In stable conditions, the extracted RFF achieves over 99% accuracy. While the accuracy decreases in more changing environments, we still obtain up to 76% accuracy in untrained locations.

Tracking UWB Devices Through Radio Frequency Fingerprinting Is Possible

TL;DR

This work investigates whether Radio Frequency Fingerprinting can be applied to Ultra-Wideband signals for device identification. By collecting a large, controlled dataset of 1.5 million+ measurements using 14 UWB boards and applying an open-set re-identification framework with CNN and Vision Transformer (ViT) models, the study demonstrates near-perfect accuracy in stable, fixed setups and meaningful generalization challenges when environments or devices vary. The authors introduce ArcFace loss within the ViT architecture to improve discriminability and show that multi-sample inputs and open-set evaluation can enhance robustness, while noting significant generalization gaps across days, locations, and unknown devices. The contributions include a pioneering UWB RFF study, a realistic dataset with controlled variation, and an enhanced DL pipeline with empirical results, highlighting both the security potential and privacy concerns of hardware-based fingerprinting in UWB systems.

Abstract

Ultra-wideband (UWB) is a state-of-the-art technology designed for applications requiring centimeter-level localization. Its widespread adoption by smartphone manufacturer naturally raises security and privacy concerns. Successfully implementing Radio Frequency Fingerprinting (RFF) to UWB could enable physical layer security, but might also allow undesired tracking of the devices. The scope of this paper is to explore the feasibility of applying RFF to UWB and investigates how well this technique generalizes across different environments. We collected a realistic dataset using off-the-shelf UWB devices with controlled variation in device positioning. Moreover, we developed an improved deep learning pipeline to extract the hardware signature from the signal data. In stable conditions, the extracted RFF achieves over 99% accuracy. While the accuracy decreases in more changing environments, we still obtain up to 76% accuracy in untrained locations.
Paper Structure (15 sections, 5 equations, 7 figures, 2 tables)

This paper contains 15 sections, 5 equations, 7 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related work
  3. Contributions
  4. Experimental setup and Dataset
  5. Setup
  6. Signal pre-processing
  7. Dataset
  8. Approach
  9. Re-identification system in Open Set problem
  10. Deep Learning models
  11. experimental results
  12. Metrics
  13. Discussions
  14. Limitation
  15. Conclusion and future work

Figures (7)

  • Figure 1: Experimental setup. UWB boards clipped in the 3D-printed mount rotated by the TurtleBot
  • Figure 2: Pipeline of the RFF extraction system through representation learning
  • Figure 3: The evaluation scenarios with their degree of complexity
  • Figure 4: t-SNE visualisation of the data before and after the projection of the DL model in the RFF feature space.
  • Figure 5: ViT architecture with parameters
  • ...and 2 more figures