Jamming Detection in MIMO-OFDM ISAC Systems Using Variational Autoencoders

TL;DR

This paper proposes an unsupervised jamming detection method for MIMO-OFDM ISAC systems using variational autoencoders to learn normal signal patterns and identify anomalies caused by jamming.

Contribution

It introduces a VAE-based framework for jamming detection in MIMO-OFDM ISAC systems, leveraging unsupervised learning on real-world echo signals.

Findings

Effective detection of jamming signals in MIMO-OFDM ISAC systems.

Outperforms conventional autoencoders in detection accuracy.

Applicable in 5G wireless network scenarios.

Abstract

This paper introduces a novel unsupervised jamming detection framework designed specifically for monostatic multiple-input multiple-output (MIMO)-orthogonal frequency-division multiplexing (OFDM) radar systems. The framework leverages echo signals captured at the base station (BS) and employs the latent data representation learning capability of variational autoencoders (VAEs). The VAE-based detector is trained on echo signals received from a real target in the absence of jamming, enabling it to learn an optimal latent representation of normal network operation. During testing, in the presence of a jammer, the detector identifies anomalous signals by their inability to conform to the learned latent space. We assess the performance of the proposed method in a typical integrated sensing and communication (ISAC)-enabled 5G wireless network, even comparing it with a conventional autoencoder.