电子学报 ›› 2022, Vol. 50 ›› Issue (6): 1370-1380.DOI: 10.12263/DZXB.20210841

所属专题: 电磁频谱智能+

• 电磁频谱智能+ • 上一篇    下一篇

分布式协作频谱感知网络中恶意节点检测和定位方法研究

吴晓晓1,2, 李刚强3, 张胜利1,2   

  1. 1.深圳大学电子与信息工程学院,广东 深圳 518060
    2.鹏城实验室,广东 深圳 518055
    3.黄淮学院信息工程学院,河南 驻马店 463000
  • 收稿日期:2021-07-05 修回日期:2022-01-31 出版日期:2022-06-25
    • 通讯作者:
    • 张胜利
    • 作者简介:
    • 吴晓晓 女,1982年出生,湖北鄂州人.深圳大学助理教授.主要研究方向为社交网络中的数据挖掘算法、5G通信网络关键技术研究、信道编码理论等.
      李刚强 男,1989年出生,河南驻马店人.黄淮学院讲师.主要研究方向为社交网络中的数据挖掘算法、分布式协议、机器学习等.
      张胜利(通讯作者) 男,1978年出生,河北人.深圳大学教授,博士生导师.主要研究方向为无线通信、区块链关键技术、物理层网络编码等. E-mail: zsl@szu.edu.cn
    • 基金资助:
    • 国家自然科学基金 (62171291); 广东省自然科学基金面上项目 (2021A1515011915); 河南省科技攻关项目 (212102210142)

Detection and Localization of Malicious Nodes in Distributed Cooperative Spectrum Sensing Network

WU Xiao-xiao1,2, LI Gang-qiang3, ZHANG Sheng-li1,2   

  1. 1.College of Electronics and Information Engineering, Shenzhen University, Shenzhen, Guangdong 518060, China
    2.Peng Cheng Laboratory, Shenzhen, Guangdong 518055, China
    3.College of Information Engineering, Huanghuai University, Zhumadian, Henan 463000, China
  • Received:2021-07-05 Revised:2022-01-31 Online:2022-06-25 Published:2022-06-25
    • Corresponding author:
    • ZHANG Sheng-li
    • Supported by:
    • National Natural Science Foundation of China (62171291); General Program of National Natural Science Foundation of Guangdong Province, China (2021A1515011915); Science and Technology Research and Development Program of Henan Province (212102210142)

摘要:

认知无线电是解决无线通信能量有效性问题的关键技术,其中频谱感知对于提高频谱的利用效率有着重要意义.针对基于共识的分布式协作频谱感知算法易受到恶意节点数据注入攻击,影响认知网络性能的问题,本文提出了两种基于神经网络的恶意节点检测和定位方法抵制网络内的恶意攻击行为,并采用基于Gossip Learning的联合学习策略进一步增强训练邻域检测和定位模型的鲁棒性.本文在9个认知节点的曼哈顿网络上模拟了分布式频谱感知的过程,并验证所提出方法的有效性.结果表明,所提方法具有良好的恶意节点检测和定位性能,联合学习策略能够使神经网络在样本局部有限的情况下学习到更多的攻击特征,提高本地检测和定位模型的可靠性.

关键词: 认知无线电, 协作频谱感知, 共识算法, 恶意节点, 神经网络, 联合学习

Abstract:

Cognitive radio is a key technology to solve the problem of energy efficiency in wireless communication, and spectrum sensing is of great significance for improving the efficiency of spectrum utilization. To solve the problem that the consensus-based distributed cooperative spectrum sensing algorithm is vulnerable to malicious node data injection attacks, we propose two approaches for detecting and localizing malicious nodes based on neural networks. And a collaborative peer-to-peer machine learning protocol(Gossip Learning) is adopted to facilitate training these neural network models. We simulate the process of distributed cooperative spectrum sensing on a 9-node Manhattan network, and verify the effectiveness of the proposed approaches. Numerical results illustrate that the proposed neural network-based approaches can effectively improve the performance of detecting and localizing malicious nodes. The collaborative learning strategy can enable nodes to learn more attack characteristics, and thus make the network more robust to attacks.

Key words: cognitive radio, cooperative spectrum sensing, consensus algorithm, malicious node, neural network, collaborative learning

中图分类号: