Digital Walkie-Talkie Identification scheme based on Sparse Representation with Multiple features

Abstract

Radio frequency fingerprinting is important in electronic warfare or internet of things network for the security. It gives allies useful information on enemy forces or prevents access of malicious nodes to the system by identifying if the signal is from the specific transmitter. In the past, RF fingerprinting was studied as searching for a unique feature or a simple classification method only. As typical signal parts for RF fingerprinting, the transient signal, the sync signal, and the error in the In-phase/Quadrature domain have been used with the machine learning algorithm such as k-nearest neighbors or support vector machine. From the last year, RF fingerprinting method with multiple features has been researched. However, the studies showing the difference of schemes on using unique feature and multiple features, have not been published. Also, there is no research on RF fingerprinting, which sparse representation based classification (SRC) algorithm is applied to, with multiple features. SRC is a qualified algorithm for the image or signal classification. In this paper, we suggest an effective RF fingerprinting method. The proposed method is to use SRC with multiple features in a signal burst. First, we used the rising transient signal, the falling transient signal, and the sync signal. Then, main lobe is extracted from the each signal. The multiple features are concatenated to use them simultaneously. We show that the classification result of our method recorded accuracy rate over 98% with three features and the result is higher, compared with convolutional neural network method.