Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Adversarial-Resilient RF Fingerprinting: A CNN-GAN Framework for Rogue Transmitter Detection

Raju Dhakal, Prashant Shekhar, Laxima Niure Kandel

TL;DR

This work tackles rogue device detection in RF fingerprinting by combining a CNN-based classifier with GAN-generated adversarial IQ samples to simulate adversaries that mimic genuine devices. The method uses temperature-scaled softmax probabilities, $p = \mathrm{softmax}(z/T)$, and a decision threshold on $p_{\max}$ to distinguish rogue from genuine devices, while also classifying genuine devices into seven identities. Key contributions include the open-set detection capability, a data preprocessing pipeline that aggregates IQ frames to 720 samples, and a hyperparameter-tuned CNN achieving a rogue F1 of $0.9871$ at $T=2.5$ and $\theta^*=0.1987$, plus a GAN that generates realistic rogue IQs validated by a Fréchet distance of $0.0545$. The results on 11 ADALM-PLUTO SDRs show high detection accuracy (rogue: $96.7\%$, genuine: $97.6\%$) and accurate seven-class identification for genuine devices, demonstrating practical viability for rogue transmitter detection in resource-constrained wireless environments.

Abstract

Radio Frequency Fingerprinting (RFF) has evolved as an effective solution for authenticating devices by leveraging the unique imperfections in hardware components involved in the signal generation process. In this work, we propose a Convolutional Neural Network (CNN) based framework for detecting rogue devices and identifying genuine ones using softmax probability thresholding. We emulate an attack scenario in which adversaries attempt to mimic the RF characteristics of genuine devices by training a Generative Adversarial Network (GAN) using In-phase and Quadrature (IQ) samples from genuine devices. The proposed approach is verified using IQ samples collected from ten different ADALM-PLUTO Software Defined Radios (SDRs), with seven devices considered genuine, two as rogue, and one used for validation to determine the threshold.

Adversarial-Resilient RF Fingerprinting: A CNN-GAN Framework for Rogue Transmitter Detection

TL;DR

This work tackles rogue device detection in RF fingerprinting by combining a CNN-based classifier with GAN-generated adversarial IQ samples to simulate adversaries that mimic genuine devices. The method uses temperature-scaled softmax probabilities, , and a decision threshold on to distinguish rogue from genuine devices, while also classifying genuine devices into seven identities. Key contributions include the open-set detection capability, a data preprocessing pipeline that aggregates IQ frames to 720 samples, and a hyperparameter-tuned CNN achieving a rogue F1 of at and , plus a GAN that generates realistic rogue IQs validated by a Fréchet distance of . The results on 11 ADALM-PLUTO SDRs show high detection accuracy (rogue: , genuine: ) and accurate seven-class identification for genuine devices, demonstrating practical viability for rogue transmitter detection in resource-constrained wireless environments.

Abstract

Radio Frequency Fingerprinting (RFF) has evolved as an effective solution for authenticating devices by leveraging the unique imperfections in hardware components involved in the signal generation process. In this work, we propose a Convolutional Neural Network (CNN) based framework for detecting rogue devices and identifying genuine ones using softmax probability thresholding. We emulate an attack scenario in which adversaries attempt to mimic the RF characteristics of genuine devices by training a Generative Adversarial Network (GAN) using In-phase and Quadrature (IQ) samples from genuine devices. The proposed approach is verified using IQ samples collected from ten different ADALM-PLUTO Software Defined Radios (SDRs), with seven devices considered genuine, two as rogue, and one used for validation to determine the threshold.

Paper Structure

This paper contains 15 sections, 4 equations, 5 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related works
  3. Methodology
  4. Data Collection
  5. System Model
  6. CNN Architecture
  7. Threshold Calculation
  8. Hyperparameter Tuning of CNN
  9. GAN Architecture
  10. Results
  11. F1-score vs Threshold
  12. Comparison of Real and Generated I/Q Constellations
  13. Confusion matrices demonstrating the performance of the proposed approach
  14. Limitation and Future works
  15. conclusion

Figures (5)

  • Figure 1: RF Fingerprinting Pipeline using CNN and GAN
  • Figure 2: Distribution of max softmax probabilities for validation set
  • Figure 3: Real vs generated I/Q constellation
  • Figure 4: Performance of the proposed approach to distinguish between rogue and genuine devices.
  • Figure 5: Overall confusion matrix showing both the detection of genuine and rogue devices and the classification of genuine devices into their respective classes.