Research On Iot Device Identification Method

With the continuous development of the Internet of Things technology,the types and number of Internet of Things devices continue to increase.Identifying the types and brands of different Io T devices effectively is the basis for analyzing the security situation of the Internet of Things and conducting network security assessments.The current methods of identifying Io T devices have certain limitations,including limited identification types,insufficient accuracy,and poor scalability.In order to explore more effective identification methods for Io T devices,a semisupervised identification method based on clustering and a semi-supervised identification method based on multi-classifier integration are proposed in this thesis.The main work of this thesis is as follows:First,clustering algorithms and classification algorithms are introduced respectively.For clustering algorithms,the principles of four clustering algorithms are introduced,and the advantages and disadvantages of the four algorithms are compared.Based on the characteristics of Io T device information,a constraint seed k-means algorithm is selected as a clustering algorithm for Io T device information.For classification algorithms,the principles of the three algorithms of logistic regression,decision tree and support vector machine are explained.Then,by analyzing the characteristics of the three algorithms,the three algorithms are used as the basic algorithms for generating the based classifier in semi-supervised algorithm based on multi-classifier integration.Secondly,due to the problems of poor clustering and difficulty in finding new device categories using traditional clustering methods,this thesis proposes a semi-supervised identification method based on clustering.The traditional constraint seed k-means algorithm is optimized by selecting the cosine distance as the distance measure and introducing the Z-Score value for similarity measurement.The improved algorithm has higher recognition accuracy and better scalability.Afterwards,in order to solve the problem that the effect of device recognition is poor when there are limited labeled samples and many unlabeled samples,this thesis proposes a semisupervised identification method based on multi-classifier integration.In order to overcome the problems that the conventional collaborative training algorithm is difficult to generate effective views,and the iteration process introduces noise to make the identification worse,the classification performance of the proposed algorithm is ensured by constructing effective based classifiers,introducing confidence value parameters,and setting early termination conditions.Experiments show that the proposed algorithm still has high recognition accuracy under the condition of limited labeled samples.Then,the realization of the Io T device identification system is introduced.It focuses on the process of equipment information collection using Nmap and Masscan comprehensively,using feature engineering to generate feature vectors,and the implementation process of the identification module.The system provides strong support for the research and verification of identification methods.Finally,the shortcomings in this paper are summarized,and the next research directions are proposed.