基于卷积神经网络的物联网设备类型识别方法  

Classification Method of Internet of Things Device Based on Convolutional Neural Network

作　　者：董卫宇[1] 王瑞敏[1] 成辉 戚旭衍[1] DONG Weiyu;WANG Ruimin;CHENG Hui;QI Xuyan(Information Engineering University,Zhengzhou 450001,China)

机构地区：[1]信息工程大学,河南郑州450001

出　　处：《信息工程大学学报》2022年第3期257-263,共7页Journal of Information Engineering University

基　　金：国家重点研发计划资助项目(2018YFB0804503)。

摘　　要：针对传统物联网设备类型识别方法在数据特征不明显、训练数据不足的情况下,难以对设备进行准确识别的问题,提出一种基于卷积神经网络的物联网设备类型识别方法。首先从互联网上获取设备WEB接口的页面截图构造数据集,然后利用卷积神经网络的泛化能力提取截图的模糊特征,构建多标签识别模型,实现对设备类型的准确识别。与传统的基于WEB页面的设备类型识别方法相比,减少了对数据特征和规模的依赖。实验结果表明,该方法的准确率达到了78.8%,与成熟的图像识别架构Xception和ResNet50相比,更适合物联网设备类型识别。To address the problem that traditional IoT device identification methods are difficult to accurately classify devices when the data characteristics are not obvious and the training data is insufficient,a convolutional neural network based IoT device identification method is proposed.First,a screenshot of the web interface of the device is obtained from the Internet to construct a data set,and then the generalization ability of the convolutional neural network is used to extract the fuzzy features of the screenshot to build a multi-label classification model to achieve accurate device classification.Compared with traditional web-page-based device identification method,it reduces the dependence on data characteristics and scale.Simulation results show that the accuracy of this method reaches 78.8%.Compared with the mature image classification architectures like Xception and ResNet50,our method is more suitable for the classification of IoT devices.

关 键 词：截图 设备识别 机器学习 卷积神经网络 

分 类 号：TN915.08[电子电信—通信与信息系统]