Deep Learning and Optimization Approaches in an Optimized Sparse Spatial Self-Nested Graph Neural Network for Secure MU-MIMO-OFDM System: Channel Estimation, Attack Detection and Mitigation: A Review

Abstract

The increasing complexity of wireless communication systems, particularly in MU-MIMO-OFDM architectures, demands intelligent and robust solutions for channel estimation, security, and resource optimization. Deep learning techniques, especially Graph Neural Networks (GNNs), Convolutional Neural Networks (CNNs), and reinforcement learning models, have emerged as powerful tools for addressing these challenges. This paper presents a comprehensive review of deep learning and optimization approaches for secure MU-MIMO-OFDM systems, focusing on optimized sparse spatial self-nested GNN architectures. Accurate channel estimation is critical for reliable communication in MIMO-OFDM systems, where conventional methods such as least squares (LS) and minimum mean square error (MMSE) suffer from high computational complexity and limited adaptability. Deep learning-based approaches, including CNN and GNN models, have demonstrated superior performance by learning complex channel characteristics and reducing estimation errors.  Furthermore, integrating GNN with attention and sparse optimization techniques enables efficient modelling of spatial relationships among antennas and users. These models also play a crucial role in attack detection and mitigation, enhancing system security against threats such as jamming and spoofing. Despite significant advancements, challenges such as computational overhead, scalability, and data dependency persist. This paper highlights recent trends, identifies research gaps, and outlines future directions for intelligent and secure MU-MIMO-OFDM systems.