Smart Identification And Security Management Of Iot Devices Using Machine Learning

In recent years,with the development of mobile communication technology,the Internet of Things has been rapidly popularized.The heterogeneity of the devices poses a challenge to traditional Internet of Things security management.However,machine learning that are good at processing high-dimensional data have also received attention in the field of Io T security management.ML can effectively solve the difficulties in Io T security.Under such a research background,this thesis proposes machine learning-based Io T security management solutions for the perception layer,transmission layer and platform layer protection.First of all,for the protection of the perception layer,this thesis proposes a machine learning-based high-efficiency identification system for Io T devices,which includes two modules: fingerprint extraction module and device identification module.The fingerprint extraction module extracts the periodic characteristics of the flow,and generates a fingerprint matrix.Based on the importance of the extracted device fingerprint,a smart classifier like KNN can realize the device identification quickly.This thesis also proposes a coarse and fine-grained hybrid network anomaly detection system based on machine learning.The coarse-grained algorithm first detects most of the anomalous attacks.Then the feature distance is measured for fine-grained detection to obtain the local optimal solution,so as to meet the accuracy rate.Experimental results show that in the case of a lot of abnormal traffic,the anomaly detection system can still achieve good detection results,with an accuracy of more than 96%,which is far better than the detection system using a single coarse-grained algorithm.Finally,this thesis studies the application of LSTM neural network algorithm in network situation assessment for platform-level protection..Collect security parameters in the entire network through the platform layer,and extract and generate network security situation assessment indicators according to relevant principles.LSTM neural network is used to evaluate the network security situation.In the experiment,the situation assessment results are divided into three categories: excellent,good,and poor.The results show that LSTM has a significant effect on network security situation assessment.Only 4 false judgments occurred in 40 test samples.And the ROC characteristic curve performed well.The three studies conducted in this thesis can be applied to the field of identification and security management of intelligent devices in the Internet of Things in the new era.After the Internet of Things devices are connected to the network,the device identification system can be used to quickly identify the type of the device.In the traffic transmission of the Internet of Things,the coarse and fine-grained hybrid detection system can efficiently analyze the abnormal conditions in the current network and beware of the occurrence of network attacks.Network managers can assess the overall network security situation at the platform level,and provide protection and early warning for the Internet of Things based on the assessment results.