Title: Detecting 5G Signal Jammers Using Spectrograms with Supervised and Unsupervised Learning

Abstract: Cellular networks are potential targets of jamming attacks to disrupt wireless communications. Since the fifth generation (5G) of cellular networks enables mission-critical applications, such as autonomous driving or smart manufacturing, the resulting malfunctions can cause serious damage. This paper proposes to detect broadband jammers by an online classification of spectrograms. These spectrograms are computed from a stream of in-phase and quadrature (IQ) samples of 5G radio signals. We obtain these signals experimentally and describe how to design a suitable dataset for training. Based on this data, we compare two classification methods: a supervised learning model built on a basic convolutional neural network (CNN) and an unsupervised learning model based on a convolutional autoencoder (CAE). After comparing the structure of these models, their performance is assessed in terms of accuracy and computational complexity.