MMU-STCNN-BDQ: a deep reinforcement learning framework for secure and energy-efficient beamforming in 6G mMIMO...

Area of Science:

• Wireless Communications
• Artificial Intelligence
• Network Security

Background:

• 6G massive MIMO networks face challenges in energy efficiency and security, particularly in mmWave bands.
• Existing beamforming methods struggle to balance performance with these critical requirements.

Purpose of the Study:

• To develop a hybrid AI framework for energy-efficient and secure beamforming in 6G massive MIMO networks.
• To enhance adversarial robustness and reduce energy consumption in wireless communications.

Main Methods:

• Utilized a Massive Multi-User Spatiotemporal Convolutional Neural Network (MMU-STCNN) with Multi-User Attention (MUA) for secure beamforming prediction from channel state information (CSI).
• Employed Bi-Gated Deep Q-Learning (BDQ) for dynamic optimization of beamforming vectors, enhancing security and energy efficiency.
• Integrated spatiotemporal and pooling layers within the MMU-STCNN for robust feature extraction.

Main Results:

• The proposed hybrid framework significantly outperformed baseline methods (DNN, DQN, ARBF) in energy efficiency and security metrics.
• Demonstrated reductions in Mean Squared Error (MSE) and Bit Error Rate (BER), alongside increased throughput for varying user numbers and SNR levels.
• Achieved high spectral and energy efficiency, proving effectiveness in large-scale multi-user communication environments.

Conclusions:

• The hybrid MMU-STCNN and BDQ framework offers a key solution for intelligent, secure, and energy-efficient 6G mMIMO networks.
• The approach effectively addresses adversarial robustness and energy efficiency challenges in next-generation wireless systems.
• Validated effectiveness across multiple performance metrics, highlighting its suitability for future 6G deployments.