基于星座图和卷积神经网络的射频指纹识别

Radio frequency fingerprint identification based on constellation and convolutional neural network

摘要:

基于无线设备物理层的射频指纹识别是保障通信安全的有效途径。传统射频特征提取方法容易受到信道的信噪比变化的干扰，难以适应动态信噪比下的通信场景。因此，本文提出了一种基于卷积神经网络的射频指纹识别方法，实现了动态信噪比下的射频指纹识别，显著改善了低信噪比下的识别准确率。本文通过搭建实验系统对4台不同功放设备进行识别，实验结果表明，在信噪比为0.5~14.5 dB范围内，该方法的综合识别率达89.4%。

Abstract:

Radio Frequency(RF) fingerprinting identification based on the physical layer of wireless devices is an effective way to ensure communication security. The conventional RF feature extraction methods are susceptible to interference from changes in the Signal-to-Noise Ratio(SNR) of the channel, which are not suitable to dynamic SNR communication situation. A RF fingerprint identification method based on Convolutional Neural Network(CNN) is proposed, which could fulfill RF fingerprinting identification under dynamic SNR condition and significantly improve the recognition rate under low SNR condition. In addition, the experiments are implemented to identify four different power amplifier devices. The experimental results show that the comprehensive recognition rate of the proposed method is 89.4% under dynamic SNR of 0.5~14.5 dB.