A Correlation Spectrum-based Fingerprint Extraction Method for 1000BASE-T Devices

In the local area network systems, numerous terminals are connected via 1000BASE-T networks of high compatibility, while the access security of terminals keeps compromised by attacks such as mac spoofing. Device fingerprint stems from the distinctive characteristics of the device components within the physical layer, which makes it a valid proof of identity to solve such authentication problems occurring at the digital layer. However, for the variation of encoding schemes and the presence of duplex data streams, traditional device fingerprint extraction methods are not adaptable to 1000BASE-T network terminals. Therefore, this paper proposes a novel fingerprint extraction method applicable to gigabit duplex devices based on spectral features through the utilization of the uncorrelated nature of data sent by full-duplex devices on both sides of the 1000BASE-T network. In our experiment, the correlation spectrum-based fingerprint separated from bi-directional data streams are used for device identification after feature degradation, with a total classification accuracy of 99.6 % achieved by linear discriminant classifier, which demonstrates the stability and validity of the device fingerprint features presented in this paper.