A Radio Frequency Fingerprint Extraction Method Based on Cluster Center Difference.

In today's information-based society, network security is becoming more and more crucial. Access authentication is an important way to ensure network security. Radio frequency fingerprints reflect the essential characteristics of wireless devices in physical layer and are difficult to be cloned, which could achieve a reliable identification of wireless devices and enhance wireless network access security. As the existing RFECTF (Radio Fingerprint Extraction based on Constellation Trace Figure) technology has some uncertain parameters and its correct recognition rate and anti-noise performance could be further improved, a radio frequency fingerprint extraction method based on cluster center difference is proposed. Combining this method with a random forest classifier, the effectiveness of the method was verified by constructing an actual RF fingerprinting system. Experiments show that when the SNR is 15 dB, the correct recognition rate of the system can reach 97.9259%, and even reaches 99.6592% while the SNR increases to 30 dB.