A Survey of Methods and Architectures for Optimized Sparse Spatial Self-Nested Graph Neural Networks for Secure MU-MIMO-OFDM Systems: Channel Estimation, Attack Detection and Mitigation

Abstract

The increasing demand for high data rates, ultra-reliable communication, and intelligent network management in 6G wireless systems has accelerated research in MU-MIMO-OFDM technologies. However, challenges such as accurate channel estimation, interference management, and security threats remain significant barriers. Recent advancements in artificial intelligence, particularly Graph Neural Networks (GNNs), have shown promising capabilities in modelling complex wireless environments. This survey explores methods and architectures based on optimized sparse spatial self-nested GNNs for enhancing channel estimation, attack detection, and mitigation in MU-MIMO-OFDM systems. Traditional techniques such as Least Squares (LS), Minimum Mean Square Error (MMSE), and compressive sensing are compared with deep learning-based approaches including CNNs, RNNs, GANs, and GNNs. The study highlights how GNNs effectively model spatial dependencies and multi-user interference in wireless networks. Furthermore, AI-driven security frameworks are reviewed for detecting adversarial attacks such as jamming, spoofing, and pilot contamination. A comparative analysis of 30 recent studies is provided to evaluate performance improvements in accuracy, spectral efficiency, and computational complexity. The survey concludes with open challenges and future research directions, emphasizing the role of lightweight and scalable GNN architectures in next-generation 6G communication systems.