基于注意力探索网络的MIMO高斯窃听信道预编码方法研究  

Research on MIMO Gaussian Wiretap Channel Precoding Method Based on Attention Exploration Network

作　　者：车佳祺 王强 国梁 陈镜伊 CHE Jiaqi;WANG Qiang;GUO Liang;CHEN Jingyi(School of Information and Communication Engineering,Beijing University of Posts and Telecommunications,Beijing 100876,China)

机构地区：[1]北京邮电大学信息与通信工程学院,北京100876

出　　处：《移动通信》2024年第10期94-97,105,共5页Mobile Communications

基　　金：国家自然科学基金“多天线无线干扰信道的机密消息传输技术研究”(62071066)。

摘　　要：针对多输入多输出高斯窃听信道,提出了一种基于注意力探索的深度网络预编码方法。首先,在基础DNN神经网络中引入跨层注意力机制,提高网络表达能力和并行性。之后,使用传统预编码算法预先生成的数据集对该网络进行预训练。最后,将噪声项添加到该网络模型的输出层,以探索具有更高保密率的协方差矩阵并重训练该网络。仿真结果表明,与当前的方法相比,提出的方法实现了更高的保密率,并有效控制运算时间成本。This paper proposes a deep network precoding method based on attention exploration for multiple-input multiple-output(MIMO)Gaussian wiretap channels.Firstly,a cross-layer attention mechanism is introduced into the basic DNN neural network to enhance its expressive capabilities and parallelism.Then,the network is pre-trained using a dataset pre-generated by traditional precoding algorithms.Finally,the noise is added to the output layer of the network model to explore covariance matrices with higher secrecy rates and train the network.Simulation results show that compared with the current method,the method proposed in this paper achieves higher secrecy rates and effectively controls the computational time cost.

关 键 词：跨层注意力 探索机制 MIMO 窃听信道 

分 类 号：TN929.5[电子电信—通信与信息系统]