水电站终端物联网恶意设备识别方法与仿真  

Identification Method and Simulation of Malicious Devices in Internet of Things at Hydropower Station Terminals

作　　者：叶波 赵娅 王荣 邹佳成 YE Bo;ZHAO Ya;WANG Rong;ZOU Jia-cheng(Spic Guizhou Jinyuan Zunyi Hydropower Development CO.,LTD,Zunyi Guizhou 563000,China)

机构地区：[1]国家电投集团贵州金元遵义水电开发有限公司,贵州遵义563000

出　　处：《计算机仿真》2025年第1期101-105,共5页Computer Simulation

基　　金：国家电投集团贵州金元股份有限公司科技项目(KY-C-2021-JY-SD02)。

摘　　要：水电站终端物联网接入设备的数据量过大时,接入设备数据难以有效分配服务器,增加了数据传输丢包率,影响了恶意设备识别的效果。为此,提出了边缘计算下水电站终端物联网恶意设备识别方法。基于边缘计算,设计了边缘服务器数据传输方案,通过约束资源以匹配服务器,优化了物联网数据传输的安全性,降低了丢包率。采用反向提取技术改进了LSTM神经网络,设计出了BiLSTM算法,实现对边缘服务器数据的正反向特征并行提取,并将特征融合。利用融合后的特征训练支持向量机,识别恶意设备。实验结果表明,上述方法的特征提取效果较好,重要度在95%以上,丢包率变化幅度低于1%,恶意设备识别的马修斯系数在0.975以上。When the amount of data of the Internet of Things access equipment of the hydropower station terminal is too large,it is difficult for the access equipment data to effectively distribute the server,which increases the packet loss rate of data transmission and affects the effect of malicious equipment identification.Therefore,an edge computing method is proposed to identify malicious devices in the Internet of Things of hydropower station terminals.Based on edge computing,an edge server data transmission scheme is designed.By constraining resources to match the server,the security of data transmission in the Internet of Things is optimized and the packet loss rate is reduced.The LSTM neural network is improved by using reverse extraction technology,and the BiLSTM algorithm is designed to achieve parallel extraction of forward and reverse features of edge server data,and feature fusion.The fused features are used to train the support vector machine to identify malicious devices.The experimental results show that the feature extraction effect of the above methods is good,the importance is more than 95%,the packet loss rate is less than 1%,and the Matthews coefficient of malicious device recognition is more than 0.975.

关 键 词：边缘计算 水电站终端物联网 恶意设备识别 支持向量机 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]