基于改进联邦学习的无线通信网络入侵安全检测方法
投稿时间:2024-01-08  修订日期:2024-04-08  点此下载全文
引用本文:
摘要点击次数: 33
全文下载次数: 0
作者单位邮编
熊磊* 汉中职业技术学院 723002
基金项目:1、陕西省中华职业教育社2023年职业教育研究课题,课题编号:ZJS202367; 2、汉中职院2021年院级教学研究与教学改革项目(提质培优行动计划项目),项目编号:HZZYJY2021006
中文摘要:针对目前无线通信网络领域存在的安全问题,提出了一种基于改进联邦学习的无线通信网络入侵安全检测方法。改进联邦学习框架也包含本地训练和服务器聚合两部分。本地训练基于注意力-门控循环单元(Attention- gated recurrent unit,AGRU)训练本地数据,并上传服务器。服务器基于注意力机制(Attention mechanism,AM)为每个客户端赋予不同的权重,并对加权后的模型参数进行聚合。实验阶段,通过消融实验验证了所提改进联邦学习框架的有效性。与FedAvg和FedNovaka联邦学习框架相比,所提方法综合性能最优。所提方法为无线通信网络入侵检测的发展提供了一定借鉴,具有广泛的应用前景。
中文关键词:文献通信网络  联邦学习  深度学习  卷积神经网络  注意力机制
 
A Wireless Communication Network Intrusion Security Detection Method Based on Improved Federated Learning
Abstract:A wireless communication network intrusion security detection method based on improved federated learning is proposed to address the current security issues in the field of wireless communication networks. Improving the federated learning framework also includes two parts: local training and server aggregation. Local training is based on Attention Gated Recurrent Unit (AGRU) to train local data and upload it to the server. The server assigns different weights to each client based on the Attention mechanism (AM) and aggregates the weighted model parameters. In the experimental stage, the effectiveness of the proposed improved federated learning framework was verified through ablation experiments. Compared with FedAvg and FedNovaka federated learning frameworks, the proposed method has the best overall performance. The proposed method provides a certain reference for the development of intrusion detection in wireless communication networks and has broad application prospects.
keywords:Literature communication network  Federated learning  Deep learning  Convolutional neural networks  Attention mechanism
查看全文   查看/发表评论   下载pdf阅读器