基于LSTM模型的无线物联网设备识别方法  

Method of Wireless IoT Devices Identification Based on LSTM Model

作　　者：郑熠[1] 田辉 马茜[1] ZHENG Yi;TIAN Hui;MA Qian(The 54th Research Institute of CETC,Shijiazhuang 050081,China;China Mobile Communications Group Hebei Co.,Ltd.,Shijiazhuang 050035,China)

机构地区：[1]中国电子科技集团公司第五十四研究所,河北石家庄050081 [2]中国移动通信集团河北有限公司,河北石家庄050035

出　　处：《无线电通信技术》2024年第3期597-602,共6页Radio Communications Technology

摘　　要：随着无线物联网(Internet of Things,IoT)业务的兴起,海量设备的接入,无线网络受攻击的可能性大大增加,无线IoT设备的安全问题越来越重要。提出了一个基于深度机器学习长短期记忆(Long Short-Term Memory,LSTM)模型的无线IoT设备识别方法,用于甄别非法入侵的设备或者发现已经被入侵后通信异常的设备。所提方法的创新点在于通过深度机器学习对IoT设备公开传输的帧头信息进行分析识别,而不必深入分析承载信息,不依赖于易被修改和伪装的IP地址等身份信息,因此不受通信信息加密的影响,也不受各类伪装地址及其他入侵方法的影响。所提方法的应用可以自动快速地识别出非授权设备或者被入侵的授权设备,更好地保障网络安全。With the rise of wireless Internet of Things(IoT)services,a massive number of devices are being connected to networks.More,wireless networks are more susceptible to external attacks,making security issues for wireless IoT devices increasingly important.Wireless IoT device identification proposes a wireless IoT device identification method based on deep machine learning Long Short-Term Memory(LSTM)model,which can be used to identify illegally invaded devices or to discover devices with abnormal communication after being invaded.The innovation of this recognition method lies in the analysis and identification of the frame header information publicly transmitted by IoT devices through deep machine learning,without in-depth analysis of the bearer information,and does not rely on the identity information such as IP addresses that are easy to be modified and disguised.So it is not affected by the encryption of communication information,nor is it affected by various intrusion methods such as disguised addresses.

关 键 词：网络安全 深度机器学习 无线物联网 时间序列 长短期记忆模型 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]