Primary-User Emulation Detection Using Database-Assisted Frequency-Domain Action Recognition

Abstract

In this paper, we propose an approach for detecting primary-user emulation (PUE) attacks in cognitive radio (CR) networks based on the application of action recognition techniques in the frequency domain. Specifically, we apply this method to analyze the fast Fourier transform (FFT) sequences of wireless transmissions operating across a CR network environment and then use a relational database and an artificial neural network to classify their actions in the frequency domain. Based on the previous approach proposed by the authors, this new approach is initiated via energy detection to locate the potential PU emulators within a specific frequency band. The approach employs a relational database system to record the motion-related feature vectors of PUs on this frequency band. When an intercepted transmission does not have a match record in the database, this transmission is considered from the PUE. Otherwise, a covariance descriptor will be calculated and fed into an artificial neural network for further classification. The proposed approach is validated via computer simulations and by experimental hardware implementations using a software-defined radio (SDR) platform. The computer simulations show that our new approach is more efficient than the authors' previous approach when there are multiple PUs in the network. The hardware experiment shows that the proposed approach can maintain system performance in terms of percentage of correct classification.