Proceedings of the 13th EAI International Conference on Mobile Multimedia Communications, Mobimedia 2020, 27-28 August 2020, Cyberspace

Research Article

Radio Frequency Fingerprint Identification Based on Multi-Intervals Differential Constellation Trace Figures

Download901 downloads
  • @INPROCEEDINGS{10.4108/eai.27-8-2020.2294331,
        author={yang  yang and Aiqun  Hu and jiabao  yu},
        title={Radio Frequency Fingerprint Identification Based on Multi-Intervals Differential Constellation Trace Figures},
        proceedings={Proceedings of the 13th EAI International Conference on Mobile Multimedia Communications, Mobimedia 2020, 27-28 August 2020, Cyberspace},
        publisher={EAI},
        proceedings_a={MOBIMEDIA},
        year={2020},
        month={11},
        keywords={radio frequence fingerprint differential constellation trace figures 3d-2d cnn motion features zigbee},
        doi={10.4108/eai.27-8-2020.2294331}
    }
    
  • yang yang
    Aiqun Hu
    jiabao yu
    Year: 2020
    Radio Frequency Fingerprint Identification Based on Multi-Intervals Differential Constellation Trace Figures
    MOBIMEDIA
    EAI
    DOI: 10.4108/eai.27-8-2020.2294331
yang yang1, Aiqun Hu2,*, jiabao yu2
  • 1: southeast university
  • 2: Southeast University
*Contact email: aqhu@seu.edu.cn

Abstract

Differential constellation trace figure (DCTF) has been demonstrated good performance on radio frequency fingerprint (RFF) identification. However, DCTF easily blurs at low SNRs. This paper proposes two novel RFF identification methods for ZigBee devices based on multi-intervals DCTFs. First, a low-complexity motion features extraction method is devised based on high-density regions. Besides, an improved 3D-2D CNN model is utilized to extract motion features and spatial features. We collected 54 different ZigBee devices for experiments and classified them by these two methods. The experimental results show that compared with using single DCTF, which is generated by a single differential interval, these two methods can effectively improve the identification accuracy at different SNR levels. The classification accuracy rate of the 3D-2D CNN classifier is over 92% even under the SNR level of 5 dB.