Research And Implementation Of User Monitoring Mechanism Based On Multidimensional Characteristics In Identifier-Based Network

With the continuous expansion of the Internet,network services and network users have shown explosive growth.The illegal acts of using network platforms to commit crimes are also increasing day by day.Criminal acts such as network fraud have caused negative social effects and increased social instability factors.From the point of network security,the network is required to achieve fine-grained and refined management of users.However,the current user monitoring mechanisms are mostly limited by the complex data transmission mode in the network,and can not meet the needs of the fine-grained man-agement.Therefore,in order to meet the requirments of the further development of infor-mation networks and realize the fine-grained management of users,this thesis designs and implements a user monitoring system with multidimensional characteristics based on the Identifier-based network.The system can accurately describe user characteristics in fine-grained,and provide user-level data support for the realization of the fine-grained controllability of the entire network.The multidimensional features user monitoring sys-tem proposed in this thesis obtains the basic information of users through the access au-thentication of network users,and then obtains the user behavior information by analyz-ing the network data of users.From these two perspectives,the user identity information is stored structurally to generate the user identity tag information,which is used by net-work access control decision makers to formulate access control policies.The specific work is as follows.(1)This thesis summarizes the status of the user identity management and the user behavior analysis,designs and implements a user monitoring system based on multidi-mensional features in the Identifier-based network,combining with large data scenarios and system requirements.This system introduces a large data platform and a computing framework,and designs a distributed user flow storage model to enable the system to carry complex and massive network data.Besides,a data parallel processing scheme is proposed,which enables the technical means and algorithm models used in the system implementation to be applied in an efficient parallel network environment.(2)User information is collected from both static and dynamic perspectives,user attributes are depicted in fine granularity,and user multidimensional features information is provided for the network access control model.User static information is acquired through access authentication and user registration.User dynamic information needs to be analyzed through the traffic data generated by users,focusing on the analysis of user interest and confidence.And this thesis provides an interface to communicate with the service subsystem and policy subsystem in network access control.(3)In this thesis,K-means algorithm is optimized in parallel,and a user interest clustering algorithm based on Spark platform is designed and implemented,which can be used to efficiently get user interest from user search keywords and solve the bottleneck of K-means clustering performance of single node.The convolutional neural network al-gorithm is optimized in parallel,and the network flow classification algorithm based on Spark platform is designed.At the same time,the Spark Streaming framework is used for real-time network flow classification to solve the real-time classification problem of mas-sive network flows.The experimental results show that the algorithm performs well in accuracy and real-time.