Physics-Informed Structure Anchoring With Capture-Aware Prototype Calibration for Cross-Environment RF Fingerprinting

Jul 16, 2026

Upload by Violet

Radio frequency fingerprint identification (RFFI) uses transmitter-specific hardware imperfections as a physical-layer identity cue for IoT devices, yet deep RFFI models degrade when the acquisition environment changes. In multi-antenna reception, degradation is driven by receiver-array topology, frequency-offset dynamics, and capture-dependent target structure that can distort embeddings and shift source-trained boundaries. This work introduces physics-informed structure anchoring with capture-aware prototype calibration (PISA-CAPC), separating representation anchoring from fixed-backbone target calibration and enabling label-free capture-aware score calibration.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages17

Reading Time43 min

Abstract

Introduction

Motivation: cross-environment degradation in RFFI

Problem framing: representation-level and decision-level mismatch

Proposed Framework (PISA-CAPC)

Representation stage with topology graph and CFO descriptors

Contextual residual suppression around identity representations

Deployment stage with unlabeled capture-aware prototype calibration (U-CAPC)

Experimental Results

Multi-antenna WiFi benchmark performance and transductive setting

Ablation study for complementary gains

Physics-Informed Structure Anchoring With Capture-Aware Prototype Calibration for Cross-Environment RF Fingerprinting Fengchong Yao, Jianbing Li, Qing Liu, Qikun Liu, Kefeng Song, Haitao Li, and Song Wang arXiv :2607 .09760v1 [ ee ss . SP] 6 Jul 2026 Abstract—Radio frequency fingerprint identification (RFFI) uses transmitter-specific hardware imperfections as a physicallayer identity cue for Internet of Things (IoT) devices, but deep RFFI models often degrade when the acquisition environment changes. In multi-antenna reception, this degradation is not merely a generic distribution shift. It is also shaped by receiver-array topology, frequency-offset dynamics, and capturedependent target structure, which can distort embeddings and move source-trained decision boundaries. This article proposes physics-informed structure anchoring with capture-aware prototype calibration (PISA-CAPC), a framework that separates source representation anchoring from fixed-backbone target calibration. The representation stage organizes antenna tokens with a topology graph and modulates the graph using CFO-derived acquisition-dynamics descriptors. Bounded contextual residual suppression is then applied around the identity representation. At deployment, unlabeled capture-aware prototype calibration (U-CAPC) calibrates target decision scores through capturelocal prototype evidence under a fixed representation, mitigating boundary shift without requiring target-domain backbone updates or target labels. On a measured ten-transmitter multiantenna WiFi benchmark, PISA-CAPC achieves 0.9257 targetdomain mean Macro-F1 under a balanced transductive setting. Ablations confirm that topology-guided structure anchoring, contextual residual suppression, and capture-aware calibration contribute complementary gains. These results establish PISACAPC as a fixed-backbone route to cross-environment RFFI, coupling physically motivated representation learning with labelfree, capture-aware decision calibration. Index Terms—Internet of Things, physical-layer authentication, radio frequency fingerprint identification, antenna topology, structure-anchored representation, contextual residual suppression, prototype calibration. I. INTRODUCTION RELIABLE device authentication remains difficult in open Internet of Things (IoT) deployments. Many low-cost, mobile, or noncooperative devices operate at large scale under limited computational and communication resources. Credential-based mechanisms alone do not fully cover this setting: credentials can be copied, lost, or unavailable to the receiver during passive monitoring. Radio frequency fingerprint identification (RFFI), also known as specific emitter identification (SEI), offers a complementary physical-layer identity cue Fengchong Yao, Jianbing Li, Qing Liu, Qikun Liu, Kefeng Song, Haitao Li, and Song Wang are with the School of Information Systems Engineering, Information Engineering University, Zhengzhou 450001, China. E-mail: Fengchong Yao ([phoenixly@126.com](phoenixly@126.com)), Jianbing Li ([li_jb@126.com](li_jb@126.com)), Qing Liu ([liuqing8123@163.com](liuqing8123@163.com)), Qikun Liu ([ed-liuqikun@163.com](ed-liuqikun@163.com)), Kefeng Song ([annx1990@163.com](annx1990@163.com)), Haitao Li ([lihaitao_01@163.com](lihaitao_01@163.com)), Song Wang ([wangsong8190@163.com](wangsong8190@163.com)). Corresponding author: Jianbing Li (e-mail: [li_jb@126.com](li_jb@126.com)). by exploiting small and uncontrollable hardware differences caused by manufacturing tolerances, oscillator imperfections, power-amplifier nonlinearities, and other transmitter-side impairments [1],[2] . Since these fingerprints are embedded in the emitted waveform, RFFI can provide device identity evidence without modifying the upper-layer communication stack. Recent deep-learning-based RFFI methods have shown strong recognition ability when training and testing samples are acquired under comparable conditions. Convolutional network

FAQs

Why do deep RFFI models degrade under environment changes?

Because receiver-array topology, frequency-offset dynamics, and capture-dependent target structure distort the learned embeddings and can shift decision boundaries, making source-domain-trained features unreliable in new acquisition contexts.

What does PISA-CAPC do differently from fixed-label or purely statistical adaptation?

It anchors the source representation using physics-informed structure (topology graph and CFO-derived descriptors), then applies capture-aware prototype calibration at deployment with a fixed backbone to mitigate boundary shift without target-domain backbone updates or target labels.

How is unlabeled capture-aware prototype calibration (U-CAPC) applied during deployment?

U-CAPC calibrates target decision scores using capture-local prototype evidence under a fixed representation, reducing boundary shift while avoiding reliance on target labels or backbone retraining.

Jul 16, 2026

Upload by Violet