Deep convolutional fennec fox and sparse spatial self-nested graph neural networks for secure channel estimation and attack mitigation for MU-MIMO-OFDM

Abstract

In general, a Multi-User Multiple Input Multiple Output Orthogonal Frequency Division Multiplexing system (MU-MIMOOFDM) can support a multiplicity of users to access a given base station through the use of multiple antennas and OFDM modulation. Nevertheless, new problems such as power management and delay optimization come into play. 5 G, 6 G, and many upcoming technologies are in the pipeline, and getting the data transmission safe is always important, and Channel Estimation is an essential element in it. Based on the proposed theoretical model of MU-MIMOOFDM, this paper designs a new Sparse Pyramid Fennec fox Fuzzy Framework (SPFFF) to enhance the comprehensive utilization of integration techniques of high-performance methods on channels; attack detection, power distribution, and delay-information optimized scheduling. This presents a Sparse Spatial Self-Nested Graph Neural Network (3SNGN) for channel estimation and attack detection since it deals with complex spatial relations and hierarchical dependencies for robust and accurate solutions. To assign power, a Deep Convolutional Pyramid-Dilated Neural Network (DCPDN) is designed to view multi-scale features and optimize the resource power. Make predictions on such data using two types of neural networks optimized by applying Fennec Fox Optimization (FFO) to improve prediction and computational performance. Newly, authors proposed a Fuzzy-based Alpine Skiing Priority (FASP) scheduling algorithm by integrating fuzzy logic with the Alpine Skiing Optimization algorithm to schedule users based on waiting time to avoid inter-user interference (IUI) and minimize the delay performance. In MATLAB, we have used the proposed method for BER (0.00012), MSE (0.00023), NMSE (0.001), and PSNR (45 dB) in the MU-MIMOOFDM system, and it outperforms the present technique. The proposed approach has also attained higher spectral capacity, a higher fairness index, 17.45% improved energy efficiency, and 12.05% lower processing time than other conventional approaches. Simulation results have proved that the proposed framework remarkably enhances the spectral efficiency while reducing latency, enhancing its robustness against attacks.