High Power Laser and Particle Beams, Volume. 36, Issue 4, 043019(2024)

Visual analysis method for RF fingerprint based on important region localization and masking

Wenbin Liu1,2, Pingzhi Fan1, Jiahuang Yang3, Yukai Li3, Yuhao Wang3, and Hua Meng3、*
Author Affiliations
  • 1School of Information Science & Technology, Southwest Jiaotong University, Chengdu 611756, China
  • 2The 30th Research Institute of China Electronics Technology Group Corporation, Chengdu 610041, China
  • 3School of Mathematics, Southwest Jiaotong University, Chengdu 611756, China
  • show less
    Figures & Tables(8)
    Network architecture of MiddleResNet and LittleResNet
    Comparison of time-domain pulse signal heat maps between MiddleResNet and LittleResNet
    Important areas of emitters when using MiddleResNe and LittleResNet
    Test results by continuous mask and different samples comparison
    • Table 1. Performance comparison results of three models

      View table
      View in Article

      Table 1. Performance comparison results of three models

      modelaccuracyparameter number/106
      5 dB9 dB14 dBall data
      LittleResNet0.76050.87750.96050.88120.153
      MiddleResNet0.76650.89900.96850.89300.254
      Transformer0.86700.94930.9950.94391.021
    • Table 2. Number and sequence of mispredicted samples

      View table
      View in Article

      Table 2. Number and sequence of mispredicted samples

      emitterTransformerMiddleResNetLittleResNet
      1#1(sample23)4(sample3,5,12,17)4(sample5,10, 12, 23)
      2#07(sample3,5,7,14,19,22,28)6(sample3,6,7,19,22,28)
      3#000
      4#1(sample19)2(sample7,19)2(sample7,19)
      5#1(sample9)1(sample14)0
      6#000
      7#000
      8#000
      9#000
      10#000
    • Table 3. Prediction results when masking 3# and using the MiddleResNet model

      View table
      View in Article

      Table 3. Prediction results when masking 3# and using the MiddleResNet model

      sample123456789101112131415161718192021222324252627282930
      no masking333333333333333333333333333333
      1st masking332333331331333333333333232333
      2nd masking332333333333333333333333332333
      3rd masking333333333333333333333323332333
    • Table 4. Number of mispredicted samples and emitter(s) mainly mispredicted to when masking different times

      View table
      View in Article

      Table 4. Number of mispredicted samples and emitter(s) mainly mispredicted to when masking different times

      emitterMiddleResNetLittleResNet
      no masking1stmasking2ndmasking3rdmasking emitter(s) mainly mispredicted tono masking1stmasking2ndmasking3rdmasking emitter(s) mainly mispredicted to
      1#42729292#41010102#
      2#71619191#、3#61014101#、3#
      3#05222#、1#0000——
      4#21327293#21216103#、2#、5#、1#
      5#12029293#、4#0514173#
      6#01328307#、10#01523257#、10#
      7#0791610#、3#011710#、3#
      8#09202810#、7#、9#、3#07131610#、7#、6#、9#
      9#08141710#、7#、3#06131610#
      10#0211157#、3#02347#
    Tools

    Get Citation

    Copy Citation Text

    Wenbin Liu, Pingzhi Fan, Jiahuang Yang, Yukai Li, Yuhao Wang, Hua Meng. Visual analysis method for RF fingerprint based on important region localization and masking[J]. High Power Laser and Particle Beams, 2024, 36(4): 043019

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category:

    Received: Oct. 30, 2023

    Accepted: Jan. 22, 2024

    Published Online: Apr. 22, 2024

    The Author Email: Hua Meng (menghua@swjtu.edu.cn)

    DOI:10.11884/HPLPB202436.230380

    Topics