A 1000Base-T Physical Layer Fingerprint Extraction and Identification System

In 1000Base-T Ethernet, terminal access problems are often ignored. This paper proposes a novel method of extracting 1000Base-T physical layer fingerprint and builds a fingerprint identification system. It can prevent the network from being attacked by media access control (MAC) spoofing. The fingerprint is extracted from the signal characteristics of network devices so it is hard to counterfeit. The single-end signal is calculated by the pretrained convolutional neural network so we can extract the spectrum of the single-end signal. The fingerprint is extracted from the spectrum of the single-end signal and is classified after the feature extraction. In the classification and identification experiments on 8 devices, we achieve an accuracy of 85.9% on multi-classification and a high accuracy on binary-classification. This method can be used to enhance the security of wired networks, especially wired Internet of Things networks.