Intelligent prediction algorithm of secure transmission performance of covert communication systems based on GTM-Net
CHENG Hanbing
FANG Jun
LI Tiancheng
XU Lingwei
Abstract：Due to the broadcast and open nature of channels in wireless communication systems,the issue of protecting the privacy and security of wireless communication users faces great challenges.Covert com-munication can better protect the hidden information by hiding the transmission process on the transmis-sion medium to prevent the third party from detecting the communication behavior.However the mobile communication environment is complex and changeable,and the secure transmission performance is affect-ed by the complex and changeable environment,which makes it difficult to make real-time accurate predic-tion.Therefore,this paper proposes an intelligent prediction algorithm for secure transmission perform-ance of covert communication system based on GTM-Net neural network.First,a covert communication system assisted by Intelligent Reconfigurable Surface and Artificial Noise is established under N-Nakagami channel,and the secure transmission performance is analyzed by the link signal-to-noise ratio.In order to predict the secure transmission performance in real time,the GTM-Net model is designed in this paper.The model adopts a two-branch structure,one branch is a two-layer sampling-based Graph Convolutional Network for local random sampling,and a layer of Transformer is added after each layer of GraphSAGE for global feature extraction,and the other branch is a feature interaction through MLP-Mixer network.The experimental results show that,compared to other algorithms,the GTM-Net algorithm has better prediction performance.In terms of prediction accuracy,GTM-Net improves 85.36%over Graph Convo-lutional Network.
Keywords：covert communicationintelligent reflective surfacesecure transmissionintelligent predic-tionGraph Convolutional Network
Publication Date：2026-08-25
Online Publishing Date：2026-08-28(First online date of this platform, not the publication date of the document)
Pages：10( 518-527 )