电子学报 ›› 2021, Vol. 49 ›› Issue (9): 1852-1856.DOI: 10.12263/DZXB.20201340

• 科研通信 • 上一篇    下一篇

基于集成学习的全双工中继系统安全中继选择方案研究

张梦, 郑建宏, 刘香燕, 何云   

  1. 重庆邮电大学通信与信息工程学院,重庆 400065
  • 收稿日期:2020-11-26 修回日期:2021-05-17 出版日期:2021-09-25 发布日期:2021-09-25
  • 作者简介:张 梦 女,1988年10月出生于重庆市. 现为重庆邮电大学博士研究生. 主要研究方向为物理层安全技术. E-mail:mzhang@stu.cqupt.edu.cn
    郑建宏 男,1961年8月出生于四川广安市. 现为重庆邮电大学教授. 主要研究方向为无线通信技术. E-mail:zhengjh@cqupt.edu.cn
    刘香燕 女,1992年2月出生于四川资阳市. 现为重庆邮电大学博士研究生. 主要研究方向为资源分配管理.E-mail:xiangyan.leo@gmail.com
    何 云 女,1979年11月出生于湖北武汉市. 现为重庆邮电大学博士研究生. 主要研究方向为新一代宽带无线通信系统. E-mail:yunhe@cqupt.edu.cn
  • 基金资助:
    国家科技重大专项(2018ZX03001026-002);重庆市教委科技项目(KJQN201800618)

Ensemble Learning-Based Relay Selection Scheme in Full-Duplex Relay System for Secure Transmission

Meng ZHANG, Jian-hong ZHENG, Xiang-yan LIU, Yun HE   

  1. School of Communication and Information Engineering,Chongqing University of Posts and Telecommunications,Chongqing 400065,China
  • Received:2020-11-26 Revised:2021-05-17 Online:2021-09-25 Published:2021-09-25

摘要:

当处于无线信道状态信息(Channel State Information, CSI)快速变化或者多跳中继等应用场景时,利用集成学习算法解决安全中继选择问题能减少实时处理时延及计算复杂度.将合法信道和窃听信道的CSI作为训练模型输入,使系统安全容量到达最大的中继节点索引作为输出,把全双工中继系统安全中继选择问题转化为一个多类分类问题,并利用随机森林(Random Forest, RF)算法求解.安全中继选择方案的实现分为数据准备、模型建立和结果预测三个阶段.在数据准备阶段,由于RF算法要求训练模型输入为离散值,给出了均匀量化和非均匀量化两种特征提取法将CSI转化为离散值.最后,通过仿真实验验证方案性能.

关键词: 全双工, 中继选择, 物理层安全, 机器学习, 集成学习, 随机森林

Abstract:

Using the ensemble learning algorithm to solve the secure relay selection problem can reduce the real-time processing delay and computational complexity, when the wireless systems that with fast-changing channel state information (CSI) or multi-hop relay situation. The secure relay selection problem of a full-duplex relay system is modeled as a multi-class classification problem, which is solved by the random forest (RF) algorithm. The CSI of the legitimate channel and eavesdropping channel is taken as the input of the training model, and the index of the relay node which can maximize the system security capacity is taken as the output of the training model. The implementation of the proposed scheme is divided into three phases: training data preparation, model building and result prediction. In the training data preparation phase, as the input features demanded by the RF algorithm are discrete values, feature extraction methods based on both uniform and non-uniform quantization algorithms are proposed to transform the CSI into finite discrete values. Finally, simulation experiments are conducted to verify the performance of the proposed scheme.

Key words: full-duplex, relay selection, physical layer security, machine learning, ensemble learning, random forest

中图分类号: