Automatic Jammer Signal Classification Using Deep Learning in the Spectrum of AI-Enabled CR-IoT

Abstract

The emerging Internet of things (IoT) technology facilitates ubiquitous and seamless connectivity of various objects to provide different services. It is envisioned to incorporate self-awareness (SA) capabilities into the IoT devices to make the entire network autonomous and intelligent, giving the concept of cognitive radio (CR) CR-IoT network. Like other wireless networks, CR-IoT suffers from various kinds of abnormal attacks. However, due to the developments of deep learning models, it has become possible to efficiently recognize and classify malicious signals present in the signal transmission. In this work, we implemented deep learning models (AlexNet and GoogLeNet) to classify jammer signals present in a CR-IoT network using fast Fourier transform (FFT) and continuous wavelet transform (CWT) features extracted from the received orthogonal frequency division multiplexing (OFDM) signal spectrum. The CR-IoT network is considered in which users and a jammer are present. Both models are capable of classifying signals into the normal signal spectrum, jammer with high power, and jammer with low power. The performance of the proposed method is evaluated using receiver operating characteristic (ROC) curves.