Enhancing 6G network security: GANs for pilot contamination attack detection in massive MIMO systems

Abstract

Accurate channel estimation at the base station (BS) is crucial for reliable, high-throughput communication in massive multiple input multiple output (MIMO) systems. However, the process of estimating channel matrices in massive MIMO systems is prone to pilot contamination attacks. These attacks not only degrade the efficiency of channel estimation but also increase the probability of information leakage. In this paper, a method is being proposed for detecting pilot contamination signals using a deep learning model, namely a generative adversarial network (GAN), at the BS of a massive MIMO system. GAN comprises two interconnected models, the first model is the generator, which creates synthetic samples, and the second model is the discriminator, which distinguishes between real and generator-generated samples. Residual block-based convolutional neural networks (CNNs) like ResNet and reverse ResNet are implemented as discriminator model of GAN for improving the detection of pilot-contaminated signals. The generator, structured as an autoencoder, generates artificial data resembling real data from noisy signals. The simulation results demonstrate that the proposed model effectively identifies the pilot-contaminated attacks; this effectiveness is evaluated through performance matrices called accuracy and loss function. Furthermore, the performance of the model is observed by varying the number of BS antennas and pilot lengths.