| 引用本文： |
-
祁欣妤,李古月,谢仁杰,顾啸林,郭永安.面向快时变信道的车联网终端指纹识别研究[J].信息安全学报,已采用 [点击复制]
- QI Xinyu,LI Guyue,XIE Renjie,GU Xiaolin,GUO Yongan.Radio Frequency Fingerprinting for V2X Facing Rapidly Time-varying Channels[J].Journal of Cyber Security,Accept [点击复制]
|
|
| 摘要: |
| 当前车联网规模化部署背景下，系统面临节点伪装等严峻安全风险。射频指纹识别作为一种极具前景的物理层认证技术，其高鲁棒认证需求却因宽带快时变信道下指纹识别性能的严重退化而难以满足。多径衰落与多普勒频移的复杂耦合，严重掩盖了信号中微弱的硬件缺陷特征。针对该问题，本文提出了一种基于频域分布差异的指纹解耦方法。首先，基于信道与硬件损伤的卷积模型，定量分析了射频指纹、多径效应及多普勒效应在频域的分布特性。在已有研究揭示的指纹与多径“快-慢”变化差异基础上，进一步指出多普勒效应呈现为准周期性的结构性快变特征，即其引起的频谱波动在频域上呈现近似等间隔的起伏形态，与随机快变的射频指纹存在显著统计差异。基于此，设计了一种二阶段指纹过滤算法。第一阶段利用倒谱滤波剔除缓变多径分量。第二阶段采用小波熵阈值方法，根据多普勒效应与指纹分量在小波域的能量分布差异将二者分离，从而实现快时变信道下设备指纹的精确解耦。为验证所提方法的有效性，本文基于商用LTE-V2X终端（CX7100）构建多场景实测系统，并在多种信道模型、信噪比及环境条件下开展了广泛的实验。实验结果表明，在跨场景、跨速度（0-90 km/h）的复杂测试条件下，所提方法取得了91.8%的平均识别准确率，表现出显著的鲁棒性。此外，该方法在设备老化与环境漂移条件下展现出优异的长期鲁棒性，性能显著优于现有基线方法，为动态车联网环境提供了一种高可靠的物理层安全解决方案。 |
| 关键词: 射频指纹 物理层安全 快时变信道 特征解耦 |
| DOI： |
| 投稿时间：2026-03-11修订日期：2026-05-14 |
| 基金项目:国家自然科学基金（62571259、62571117、62502227、62501304）；江苏省前沿技术研发计划（BF2024065）；南京邮电大学引进人才科研启动基金项目（NY225010、NY224037） |
|
| Radio Frequency Fingerprinting for V2X Facing Rapidly Time-varying Channels |
|
QI Xinyu1, LI Guyue2, XIE Renjie1, GU Xiaolin1, GUO Yongan1
|
| (1.Nanjing University of Posts and Telecommunications;2.Southeast University) |
| Abstract: |
| Under the current context of large-scale Vehicle-to-Everything (V2X) deployment, systems are confronted with severe security risks such as node spoofing. Radio Frequency Fingerprint Identification (RFFI) has emerged as a promising physical-layer authentication technique. However, meeting the demand for highly robust physical-layer authentication is hindered by the severe performance degradation of RFFI caused by broadband, rapidly time-varying channels. The complex coupling of multipath fading and Doppler shifts severely masks the subtle hardware impairments. To address this problem, this paper proposes a fingerprint decoupling method based on frequency-domain distribution differences. First, based on a convolutional model of channel and hardware impairments, the frequency-domain distribution characteristics of the RF fingerprint, multipath effect, and Doppler effect are quantitatively analyzed. Building upon prior research revealing the "fast-slow" variation distinction between fingerprints and multipath components, this paper further identifies the Doppler effect as a quasi-periodic, structural fast-varying feature. Specifically, its induced spectral fluctuations form approximately equidistant ripples, which differ significantly in statistics from the random fast-varying RF fingerprints. Accordingly to these insights, a two-stage fingerprint filtering algorithm is designed. The first stage employs cepstral filtering to eliminate slow-varying multipath components. The second stage utilizes a wavelet entropy thresholding to separate the Doppler effect from fingerprint components based on their different energy distribution characteristics in the wavelet domain, thereby achieving precise decoupling of device fingerprints in rapidly time-varying channels. To validate the effectiveness of the proposed method, a multi-scenario experimental system is constructed using commercial LTE-V2X terminals (CX7100). Extensive experiments are conducted across diverse conditions, covering various channel models, signal-to-noise ratios, and environments. Experimental results demonstrate that under complex testing conditions involving cross-scenario and cross-speed (0–90 km/h) scenarios, the proposed method achieves an average identification accuracy of 91.8%, exhibiting significant robustness. Furthermore, it exhibits significant long-term robustness against device aging and environmental drift, significantly outperforming existing baseline methods and providing a highly reliable physical-layer security solution for dynamic V2X environments. |
| Key words: radio frequency fingerprint, physical layer security, rapidly time-varying channel, feature decoupling |
