Research On Intelligent Identification Technology Of IoT Devices

With the rapid development of the Internet of Things,people are paying more and more attention to the security of Internet of Things devices.More and more attackers use device vulnerabilities to penetrate and launch network attacks.Therefore,the research on security access mechanism of IOT devices becomes more and more important,and the identification of IOT devices is the key problem to be solved.However,due to the diversity of IOT devices and their traffic characteristics,it is difficult to achieve a unified access control mechanism.The fingerprints of IOT devices make use of the high similarity of the same type of devices in software,hardware and firmware,and can be constructed through the traffic characteristics of IOT devices,so it is an important research direction to solve the problem of device identification.According to the above problems,this paper first proposes a fingerprint generation method of IOT devices based on traffic characteristics;Based on the generated fingerprint information of IOT devices,a fingerprint identification method of IOT devices based on ensemble learning is proposed.CNN and LSTM models are used to identify the types of IOT devices.The main work of the thesis is as follows::1.Proposing a method based on device traffic information for extracting fingerprint of IOT deveice.Based on the analysis of device traffic,this paper analyzes the device traffic characteristics from the message window and traffic characteristics,and proposes two ways to extract fingerprint characteristics based on message window and time window.The experimental analysis shows that the fingerprint characteristics based on message window contain time sequence information,while the fingerprint characteristics of time window count the traffic statistical characteristics over a period of time.2.Proposing a method based on ensemble learning for IOT device recognition.On the basis of obtaining the fingerprint features of the device,the device type corresponding to the fingerprint features is used as the data tag,and an integrated classification algorithm based on CNN+LSTM is proposed.The algorithm combines the advantages of CNN and LSTM feature extractor,uses CNN to extract data feature Z^CNNbased on time window traffic feature,and uses LSTM to extract data feature Z^LSTMbased on message window traffic feature.The neural network model based on CNN+LSTM is used for self-learning of fingerprint features of devices to realize IOT device recognition.Compared with CNN,LSTM and other recognition methods,the effectiveness of IOT device recognition method based on ensemble learning is verified.3.Designing and implementing an IOT device security access application based on IOS system.The security access application for access control based on device fingerprints is developed here,from the overall logic of the system,physical connection and software design three aspects of the design and implementation of the system.The system is based on the device recognition algorithm as its core architecture,which can ensure the security of access to IoT devices and provide a strong guarantee for the stable operation of the IoT.