Emitting Devices Identification Based on Variational Mode Decomposition of Captured Bluetooth Signals

Abstract

Recently, the use of communication systems is increased therefore the security networks has become more important. Radio frequency fingerprinting (RFF) is one of the communication networks security techniques based on identification of the unique features of the RF transient signals. Transient signal detection is a very important part in the classification system. If a transient signal is not detected precisely, then the features extracted from the received signal will not show the correct unique features of the transmitting devices. For this purpose, variational mode decomposition is applied to the recorded Bluetooth (BT) signal before the transient detection process. In this paper, VMD is applied before transient detection as denoysing method in RF fingerprinting of Bluetooth device identification to increase the accuracy of transient detection and the classification rate. The method has three stages: Firstly, VMD is used to decompose the recorded BT signal into a series of band- limited modes. Then, the selected modes are added to reconstruct the original BT signal. Secondly, transient detection and different statistical features are extracted from the transients such as variance, kurtosis and skewness. Finally, a linear support vector machine (LSVM) classifier is used to identify 20 mobile phones under different SNR ranges. 20 mobile phones with 150 transients each were tested in this work. Results show that the proposed VMD identification method based on the decomposition of recorded signal outperforms the existing VMD method based on decomposition of transient signal for different ranges of SNR.