Detecting 5G Signal Jammers with Autoencoders Based on Loose Observations

Fifth-generation (5G) networks are prone to jamming attacks, which are particularly dangerous in mission-critical applications such as factory automation. In this paper, we present a simple method to detect jamming attacks with machine learning techniques operating on in-phase and quadrature (IQ) modulated signals. In particular, a convolutional autoencoder (CAE) learns the structure of the clean signal to distinguish it from jammed signals in real time. This approach requires only a loose synchronization to the OFDM symbol, while equalization and decoding are not necessary. Despite its simplicity, our technique has shown high detection rates in experiments on a 5G testbed.