Variational Mode Decomposition-Based Radio Frequency Fingerprinting of Bluetooth Devices

Abstract

Radio frequency fingerprinting (RFF) is based on identification of unique features of RF transient signals emitted by radio devices. RF transient signals of radio devices are short in duration, non-stationary and nonlinear time series. This paper evaluates the performance of RF fingerprinting method based on variational mode decomposition (VMD). For this purpose, VMD is used to decompose Bluetooth (BT) transient signals into a series of band-limited modes, and then, the transient signal is reconstructed from the modes. Higher order statistical (HOS) features are extracted from the complex form of reconstructed transients. Then, Linear Support Vector Machine (LVM) classifier is used to identify BT devices. The method has been tested experimentally with BT devices of different brands, models and series. The classification performance shows that VMD based RF fingerprinting method achieves better performance (at least 8% higher) than time-frequency-energy (TFED) distribution based methods such as Hilbert-Huang Transform. This is demonstrated with the same dataset but with smaller number of features (nine features) and slightly lower (2-3 dB) SNR levels.

On the Performance of Variational Mode Decomposition-Based Radio Frequency Fingerprinting of Bluetooth Devices

“…Moreover, for performance bounds against noise, BT devices are identified at lower SNR levels (−5 to 5 dB) by employing a Linear Support Vector Machine (LSVM) classifier. It should be noted that relatively higher SNR values (5 to 25 dB) were already reported in Aghnaiya et al's work [18]. In this way, the lower SNR bounds, in which the VMD can achieve acceptable accuracy in the classification of BT devices, is explored experimentally.…”

“…Recently, VMD has been used for the first time to decompose BT transient signals into a series of band-limited modes from which transient signals are reconstructed [18]. Then, higher order statistical (HOS) features, such as skewness, kurtosis and the variance of instantaneous amplitude, frequency and phase, are extracted from the reconstructed transients.…”

“…This article, as a follow up study of Aghnaiya et al's study, is devoted to scrutinizing the performance bounds of VMD in RFF implementation [18]. Firstly, HOS features are extracted both from band-limited modes and reconstructed transient signals individually.…”

“…Recently, noise-related methodologies to generate additive white Gaussian noise (AWGN) noise to simulate channel noise, have been provided in Xie et al's studies [21,22]. However, in this study, we considered randomly varying SNR values within a range (not a single value), as this could better represent dynamic radio links of low power IoT devices, such as BT [18]. To this end, three different datasets with different SNR levels were created based on the ranges of SNR given in literature [6,7,15] to follow approximately Gaussian distribution in order to prove that the performance is evaluated with varying SNR values within the range.…”

“…Its superiority over EMD has been reported in various applications, such as the monitoring of wind turbines [14], single hop and relaying scenarios [15], fluctuation analysis [16], and pulse radar fingerprint extraction [17]. On the other hand, regarding the performance of VMD with transient signals, VMD has been successfully demonstrated with Bluetooth (BT) devices [18].…”