Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Generic Machine Learning Framework for Radio Frequency Fingerprinting

Alex Hiles, Bashar I. Ahmad

TL;DR

This paper presents a generic ML-based framework for RF fingerprinting that is emitter-type agnostic and capable of handling multiple downstream tasks (SEI, EDA, RFEC, OS-EDA). It formalizes a fingerprint head F_theta and task heads g_phi within a multi-task learning objective, and demonstrates performance on real datasets from spaceborne AIS, DMR SIGINT, and drone RF links using diverse architectures. The results show that both complex and lightweight models can achieve strong SEI/EDA/RFEC performance, including open-set and unsupervised scenarios, while remaining viable for edge deployment. Overall, the framework provides a versatile, data-driven approach to RF fingerprinting with practical ISR, SIGINT, and counter-UAS applications.

Abstract

Fingerprinting radio frequency (RF) emitters typically involves finding unique characteristics that are featured in their received signal. These fingerprints are nuanced, but sufficiently detailed, motivating the pursuit of methods that can successfully extract them. The downstream task that requires the most meticulous RF fingerprinting (RFF) is known as specific emitter identification (SEI), which entails recognising each individual transmitter. RFF and SEI have a long history, with numerous defence and civilian applications such as signal intelligence, electronic surveillance, physical-layer authentication of wireless devices, to name a few. In recent years, data-driven RFF approaches have become popular due to their ability to automatically learn intricate fingerprints. They generally deliver superior performance when compared to traditional RFF techniques that are often labour-intensive, inflexible, and only applicable to a particular emitter type or transmission scheme. In this paper, we present a generic and versatile machine learning (ML) framework for data-driven RFF with several popular downstream tasks such as SEI, data association (EDA) and RF emitter clustering (RFEC). It is emitter-type agnostic. We then demonstrate the introduced framework for several tasks using real RF datasets for spaceborne surveillance, signal intelligence and countering drones applications.

A Generic Machine Learning Framework for Radio Frequency Fingerprinting

TL;DR

This paper presents a generic ML-based framework for RF fingerprinting that is emitter-type agnostic and capable of handling multiple downstream tasks (SEI, EDA, RFEC, OS-EDA). It formalizes a fingerprint head F_theta and task heads g_phi within a multi-task learning objective, and demonstrates performance on real datasets from spaceborne AIS, DMR SIGINT, and drone RF links using diverse architectures. The results show that both complex and lightweight models can achieve strong SEI/EDA/RFEC performance, including open-set and unsupervised scenarios, while remaining viable for edge deployment. Overall, the framework provides a versatile, data-driven approach to RF fingerprinting with practical ISR, SIGINT, and counter-UAS applications.

Abstract

Fingerprinting radio frequency (RF) emitters typically involves finding unique characteristics that are featured in their received signal. These fingerprints are nuanced, but sufficiently detailed, motivating the pursuit of methods that can successfully extract them. The downstream task that requires the most meticulous RF fingerprinting (RFF) is known as specific emitter identification (SEI), which entails recognising each individual transmitter. RFF and SEI have a long history, with numerous defence and civilian applications such as signal intelligence, electronic surveillance, physical-layer authentication of wireless devices, to name a few. In recent years, data-driven RFF approaches have become popular due to their ability to automatically learn intricate fingerprints. They generally deliver superior performance when compared to traditional RFF techniques that are often labour-intensive, inflexible, and only applicable to a particular emitter type or transmission scheme. In this paper, we present a generic and versatile machine learning (ML) framework for data-driven RFF with several popular downstream tasks such as SEI, data association (EDA) and RF emitter clustering (RFEC). It is emitter-type agnostic. We then demonstrate the introduced framework for several tasks using real RF datasets for spaceborne surveillance, signal intelligence and countering drones applications.

Paper Structure

This paper contains 30 sections, 17 equations, 13 figures, 2 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Background and Overview
  3. Contributions
  4. Paper Layout
  5. Related Work
  6. Closed-set vs Open-set Tasks
  7. Input Representations
  8. Mitigating Channel and Receiver Response Effects
  9. Problem Statement and RFF Framework
  10. Single Task RF Fingerprint Learning
  11. Closed-Set Specific Emitter Identification
  12. RFF-based Emitter Data Association
  13. Unsupervised Learning for RFEC
  14. Open Set SEI and EDA
  15. Experimentation: Datasets, Models and Set-up
  16. ...and 15 more sections

Figures (13)

  • Figure 1: Block diagram of ML-based RFF-DT.
  • Figure 2: Example pipeline for handling OS-SEI.
  • Figure 3: CS-SEI test dataset evaluation: confusion matrices. Top: VGG19 (L), BCNN (M), GLFormer (R). Bottom: BILSTM (L), FCN (R)
  • Figure 4: CS-SEI Test dataset evaluation: t-SNE data embeddings. Top: VGG19 (L), BCNN (M), GLFormer (R). Bottom: BILSTM (L), FCN (R).
  • Figure 5: CS-SEI for the drone RF links test dataset. Top: t-SNE embeddings using FCN. Bottom: confusion matrix.
  • ...and 8 more figures