基于模型融合的物联网设备识别方法

打开文本图片集

中图分类号：TN911.7-34；TP391 文献标识码：A 文章编号：1004-373X（2026）07-0083-05

引用格式：，朱俊虎，，等.基于模型融合的物联网设备识别方法[J].现代电子技术，2026，49（7）：83-87.

Abstract：Inviewof the limitationsof thecurrent network deviceidentificationtechniques，particularlyin termsof the insuficienttemporalfeatureextractionandtherestrictedrecognitionacuracyofthesinglemodel，thispaperproposesanovel IoTdeviceidentificationapproachbasedonadaptivemodelfusion.Inthisapproach，thePysharktolisutilizedtoextractritical informationsuchastimestamps，portnumbers，protocol types，andpayloadsfromtraficdata.Staticfeaturesarecaptured with convolutionalneuralnetwork（CNN），whiletimestampinformationisleveragedtoenhancetherepresentationof temporaldata. Longshort-termmemory（LSTM）networksareemployedtoextracttemporal features.Anadaptivefusionmechanismis innovativelyincorporatedtodynamicallyoptimizetheweightdistributionbetweenCNNandLSTM.AStackingensemblestrategy isadoptedforfeaturefusion，ultimatelyachievingtraffcclasificationanddeviceidentification.Theidentifcationaccuracyofthe proposed methodonthepublicdatasetAaltoisimproved to97%，which demonstrates significantadvantages over theexisting methods.To sumup，this paper provides a reliable method for eficient identification of Internet of Things devices.

Keywords：IoT;device identification;feature extraction;deep learning;traffc classification; ensemble learning

0 引言

随着信息技术的飞速发展，网络设备的种类和规模持续扩大，物联网（IoT）设备已成为互联网的重要组成部分。（剩余7242字）