Research On Recognition Technology Of Intelligent Internet Of Things Devices

In the last few years,Internet of Things(Io T)technology has achieved rapid development.A surging number of Io T devices are connected to the Internet,and their applications are growing into increasingly diversified.While the Io T devices provide convenience to people,there are also some security risks.Io T device manufacturers pay more attention to the functionality and diversity of the devices and fail to update the patchs for the devices in time;Users cannot manage and configure Io T devices well due to lack of professional knowledge.These reasons all lead to the vulnerability of Io T devices.Attackers launched attacks on these vulnerable Io T devices,which had a great impact on the security of the network.The preidentification of Io T devices connected to the network can help network administrators to set appropriate security policies according to the functionality and heterogeneity of the devices.Additionally,the security configuration and management of the devices can be realized by timely patching the devices with vulnerabilities or isolating vulnerable devices.This thesis proposes two methods for intelligent Io T device identification,including end-toend Io T device identification method and unknown Io T device identification method based on multi-dimensional information search:(1)In terms of end-to-end Io T device identification method(Io T ETEI),the original communication traffic generated by Io T devices is preprocessed into sample inputs suitable for deep learning model.Then,three designed neural network models(CNN,RNN and CNN+BILSTM)are combined to achieve accurate device identification.This method is different from the traditional Io T device identification method based on manual features.Io T ETEI directly uses the neural network model to learn the spatial and temporal characteristics of the device communication traffic,such as the position relationship of internal organization structure in network communication traffic,the time sequence of data packets and the duration of network flow,etc.Io T ETEI evades the inconvenience of manually extracting features and improves the timeliness of Io T device identification.(2)With regard to the identification method of unknown Io T devices based on multidimensional information search,this thesis extracts effective information related to device identity from the device banner information obtained through active probing and the device communication traffic captured by passive listening.Based on the two named entity recognition methods of corpus and CRFModel,more accurate and comprehensive device identity information is obtained from web pages.The granularity of identification includes three types: device manufacturer,device type,and device model.(3)Finally,to demonstrate the reliability of the solution,this thesis uses public datasets and laboratory dataset to test and verify the solution from multiple metrics such as accuracy,recall,and precision.The experimental results illustrate that the end-to-end Io T devices identification method can precisely identify Io T devices under supervised learning,and the accuracy is above 97%.The unknown Io T device identification method based on multidimensional information search can precisely identify the Io T devices in the network from three levels of identification granularity.