Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Mitigating Receiver Impact on Radio Frequency Fingerprint Identification via Domain Adaptation

Liu Yang, Qiang Li, Xiaoyang Ren, Yi Fang, Shafei Wang

TL;DR

This work tackles cross-receiver radio frequency fingerprint identification (RFFI) by casting it as a domain adaptation problem: transfer a labeled model trained on a source receiver to a target receiver with unlabeled data. It derives a theoretical bound on the target risk that depends on domain discrepancy and pseudo-label accuracy, justifying a two-pronged approach of domain alignment and adaptive pseudo-labeling. The proposed method combines KL-divergence–based domain alignment (via a DV representation) with Curriculum Pseudo-Labeling and class weighting, trained in a minimax objective using a Gradient Ascent-Descent scheme. Experiments on the Wisig WiFi dataset demonstrate that the method substantially reduces receiver-induced performance gaps and outperforms strong baselines, with notable gains on challenging cross-receiver and cross-day transfers.

Abstract

Radio Frequency Fingerprint Identification (RFFI), which exploits non-ideal hardware-induced unique distortion resident in the transmit signals to identify an emitter, is emerging as a means to enhance the security of communication systems. Recently, machine learning has achieved great success in developing state-of-the-art RFFI models. However, few works consider cross-receiver RFFI problems, where the RFFI model is trained and deployed on different receivers. Due to altered receiver characteristics, direct deployment of RFFI model on a new receiver leads to significant performance degradation. To address this issue, we formulate the cross-receiver RFFI as a model adaptation problem, which adapts the trained model to unlabeled signals from a new receiver. We first develop a theoretical generalization error bound for the adaptation model. Motivated by the bound, we propose a novel method to solve the cross-receiver RFFI problem, which includes domain alignment and adaptive pseudo-labeling. The former aims at finding a feature space where both domains exhibit similar distributions, effectively reducing the domain discrepancy. Meanwhile, the latter employs a dynamic pseudo-labeling scheme to implicitly transfer the label information from the labeled receiver to the new receiver. Experimental results indicate that the proposed method can effectively mitigate the receiver impact and improve the cross-receiver RFFI performance.

Mitigating Receiver Impact on Radio Frequency Fingerprint Identification via Domain Adaptation

TL;DR

This work tackles cross-receiver radio frequency fingerprint identification (RFFI) by casting it as a domain adaptation problem: transfer a labeled model trained on a source receiver to a target receiver with unlabeled data. It derives a theoretical bound on the target risk that depends on domain discrepancy and pseudo-label accuracy, justifying a two-pronged approach of domain alignment and adaptive pseudo-labeling. The proposed method combines KL-divergence–based domain alignment (via a DV representation) with Curriculum Pseudo-Labeling and class weighting, trained in a minimax objective using a Gradient Ascent-Descent scheme. Experiments on the Wisig WiFi dataset demonstrate that the method substantially reduces receiver-induced performance gaps and outperforms strong baselines, with notable gains on challenging cross-receiver and cross-day transfers.

Abstract

Radio Frequency Fingerprint Identification (RFFI), which exploits non-ideal hardware-induced unique distortion resident in the transmit signals to identify an emitter, is emerging as a means to enhance the security of communication systems. Recently, machine learning has achieved great success in developing state-of-the-art RFFI models. However, few works consider cross-receiver RFFI problems, where the RFFI model is trained and deployed on different receivers. Due to altered receiver characteristics, direct deployment of RFFI model on a new receiver leads to significant performance degradation. To address this issue, we formulate the cross-receiver RFFI as a model adaptation problem, which adapts the trained model to unlabeled signals from a new receiver. We first develop a theoretical generalization error bound for the adaptation model. Motivated by the bound, we propose a novel method to solve the cross-receiver RFFI problem, which includes domain alignment and adaptive pseudo-labeling. The former aims at finding a feature space where both domains exhibit similar distributions, effectively reducing the domain discrepancy. Meanwhile, the latter employs a dynamic pseudo-labeling scheme to implicitly transfer the label information from the labeled receiver to the new receiver. Experimental results indicate that the proposed method can effectively mitigate the receiver impact and improve the cross-receiver RFFI performance.
Paper Structure (27 sections, 4 theorems, 29 equations, 7 figures, 4 tables, 1 algorithm)

This paper contains 27 sections, 4 theorems, 29 equations, 7 figures, 4 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. System Model and Problem Statement
  3. Theoretical Analysis and Insights
  4. Proposed Method
  5. General Description of the Proposed Method
  6. Domain Alignment
  7. Adaptive Pseudo-Labeling
  8. Curriculum Pseudo-Labeling
  9. Class Weighting
  10. Training
  11. Experiment
  12. Setups
  13. Comparison with Existing Methods
  14. Ablation Studies
  15. Hyper-parameter Sensitivity
  16. ...and 12 more sections

Key Result

Theorem 1

Let $\hat{h}$ be an optimal model for problem eq:joint_risk. Suppose that the hypothesis space $\mathcal{H}$ has a VC dimension $d$, and that $\mathcal{S}$ and $\mathcal{T}$ each has sample size $N$. Then, for any $\rho \in (0,1)$, with probability at least $1-\rho$ (over the choice of the samples), where $\epsilon^t(\hat{h}, f^t)$ and $\epsilon^t(h^\star, f^t)$ represent the expected error probab

Figures (7)

  • Figure 1: The block diagram of RFFI system.
  • Figure 2: Waveform and spectrogram of the Wisig hanna2022wisig signals.
  • Figure 3: Scenario of cross-receiver RFFI.
  • Figure 4: Overview of the proposed adaptation method. The proposed method involves domain alignment and adaptive pseudo-labeling techniques. First, the signals from the source and target domains are input into the feature extractor in order to obtain the signal features for the source and target domains, represented as $z^s$ and $z^t$ respectively. Then, on the one hand, domain alignment reduces the gap between the distributions of $z^s$ and $z^t$. On the other hand, adaptive pseudo-labeling ensures reliable, balance, and high-confidence pseudo-labeling.
  • Figure 5: The curve of target receiver data classification accuracy with weight of estimated KL divergence $\lambda$.
  • ...and 2 more figures

Theorems & Definitions (4)

  • Theorem 1
  • Theorem 2
  • Lemma 1: ben2010theory
  • Lemma 2