3D beamforming based on deep learning for secure communication in 5G and beyond wireless networks

Three-dimensional (3D) beamforming is a potential technique to enhance communication security of new generation networks such as 5G and beyond. However, it is difficult to achieve optimal beamforming due to the challenges of nonconvex optimization problem and imperfect channel state information (CSI...

Full description

Saved in:
Bibliographic Details
Main Authors: Yang, Helin, Lam, Kwok-Yan, Nie, Jiangtian, Zhao, Jun, Garg, Sahil, Xiao, Liang, Xiong, Zehui, Guizani, Mohsen
Other Authors: School of Computer Science and Engineering
Format: Article
Language:English
Published: 2022
Subjects:
Online Access:https://hdl.handle.net/10356/157422
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Nanyang Technological University
Language: English
Description
Summary:Three-dimensional (3D) beamforming is a potential technique to enhance communication security of new generation networks such as 5G and beyond. However, it is difficult to achieve optimal beamforming due to the challenges of nonconvex optimization problem and imperfect channel state information (CSI). To tackle this problem, this paper proposes a novel deep learning-based 3D beamforming scheme, where a deep neural network (DNN) is trained to optimize the beamforming design for wireless signals in order to guard against eavesdropper under the imperfect CSI. With our approach, the system is capable of training the DNN model offline, and the trained model can then be adopted to instantaneously select the 3D secure beamforming matrix for achieving the maximum secrecy rate of the system, which is measured by the signal received by eavesdroppers outside the path of the beam. Simulation results demonstrate that the proposed solution outperforms the classical deep learning algorithm and 2D beamforming solution in terms of the secrecy rate and robust performance.