Research And Application On Fine-grained IoT Device Recognition Based On Word Embedding

Since the concept of the Internet of Things(IoT)was proposed by the International Telecommunication Union,the IoT has quickly gained worldwide recognition and become the core support of the third information science and technology revolution and the fourth industrial revolution.The accelerated integration of the IoT with new technologies such as artificial intelligence,edge computing and blockchain has spawned a large number of new products,new models and new services,triggering a new wave of social,industrial and economic development.With the continuous innovation and breakthrough of low-power wide-area technologies such as the narrow-band IoT,trillions of IoT devices have been connected to the network.At the same time,IoT devices have been widely used in traditional industries such as industry and agriculture,and have gradually become an indispensable part of the development of vertical industries such as smart cities and the Internet of Vehicles.Because there is no unified access standard and certain design defects for IoT devices,the devices are extremely vulnerable to illegal attacks or unauthorized use.In recent years,high-risk vulnerabilities related to IoT devices have been frequently exposed,and various attacks against IoT devices have emerged in an endless stream.As the common vulnerability system of IoT devices is becoming more and more perfect,how to accurately and fine-grained identify IoT devices and build security policies based on this is of great significance to grasp the security situation of IoT.This thesis makes a detailed analysis of the collection process of device attribute information in the IoT,and presents a research scheme of verifiable device attribute information.Then,a fine-grained IoT device identification scheme based on identification feature words is proposed.The public network experimental results show that the recognition scheme has high accuracy and coverage rate.Finally,this thesis designs two types of application scenarios based on local area network and public network,reflecting the important role of fine-grained IoT device identification in IoT security.The specific research contents of this thesis are as follows:(1)A verifiable scheme of IoT device attribute information based on word vector is presented.On the one hand,on the basis of the existing word vector model and with the help of the search engine query results as a corpus,a classification model is defined to verify the attribute information of IoT devices.The classification accuracy of this model reached 97.2%under the training of given positive and negative samples.On the other hand,based on this classification model,an automatic collection framework of IoT device attribute information is constructed.This framework is divided into three roles:producer,middleware and consumer,which respectively meet the three tasks of collection,transmission and processing of IoT device attribute information.The actual results show that the scheme can effectively solve the redundancy and error of equipment attribute information.(2)A fine-grained IoT device identification scheme based on identification feature words is proposed.Identification feature words are defined that a vocabulary contributes a great deal to the identification of IoT devices and can associate the attribute information of itself or other devices.Firstly,by means of the TF-IDF algorithm,the identification feature words are dug from the protocol slogan information of IoT devices.Then,a matching algorithm based on inverted index is designed for IoT device identification results.Finally,the fine-grained attribute information and protocol slogan information of the identified IoT devices are input into the device fingerprint database as the device fingerprint document.In the absence of IoT device identification feature words,Simhash algorithm is used to make approximate matching for the identification results of IoT devices.In the public network experiment,the accuracy of video surveillance and industrial control Internet of Things devices reaches more than 95%.(3)An application scenario based on the fine-grained IoT device identification scheme is designed,and its effective identification scope covers both local area network and public network.Aiming at the local area network,a self-help experiment environment based on smart home equipment was built.The protocol slogan information of the device is obtained by means of automatic traffic collection.On the basis of the existing identification scheme,all the IoT devices are successfully identified,and the model coverage reaches 80%.Aiming at the public network,the visualization panel of IoT devices is constructed by using the existing localization algorithm and the developer interface of cyberspace search engine.The panel can quickly and intuitively understand the overall situation of the network space equipment,which has important guiding significance to grasp the security situation of the network space equipment.