Study Of The Host Behavior Classification Method Based On The Network Features Of Flow And Connection

The modern age is an era of the Internet, manifested in the increasing scale of the Internet, more and more network traffic can be effectively and controllably regulatory network traffic and user behavior approach, more and more network managers' attention and expectations, and how to properly use and processing of network data, extract the network features and user behavior on end- host for effective classification and identification has thus become a hot topic all scholars in universities and research institutions.However, the identification and classification of user behavior research for the end of the current host substantially in a relatively blank slateTargeted on the situation of the present host-classification research, this article analyzes the network flow characteristics in detail, constructs the model of the users' behavior spectrum, and uses it to make the extract ion and acquire the characteristics of network flow statistics, then we bring in the machine learning technology.Then,we proposed the recognition and classification method of the end hosts based on the users' spectrum and the classification method of the e nd hosts based on the characteristics of the network connection, so as to complete the recognition and classification of the end hosts and its patterns of behavior.Specific work is as follows: 1. The recognition method of the end hosts based on the flow characteristics.This article mainly elaborates the recognition method based on the characteristics of flow in the end that can be applied to small network host,we use the characterist ic data of behavior of the flow class to deal with and treat it as the input o f the recognition method,then,through our design process and the selected algorithm for the small networks, which can identify the host.At the same time we introduced the concept of users' behavior spectrum is used to process. Firstly,we introduced the bas is of the flow characteristics and the concept and the mode of the users' behavior spectrum in detail in this article,and then we introduced in detail the end hosts behavior recognition process based on the flow characteristics from the overall to local.Finally we chose the C4.5 decision tree algorithm in data mining classification algorithm to identificate end hosts.The results show that the method that introduced users' behavior spectrum and singular value decomposition can obtain better effect than traditional statistical feature recognition method.In addition,we introduced the host recognition of experimental steps and annalyzed the experimental results. 2. The classification method of the end hosts behavior based on the characteristics of the network connection. This article mainly introduces the host behavior classification method based on the characteristics of the network connection for the small network, we used the behavior characteristics of the hosts,identifying the hosts of the small network,meanwhile this is based on the concept of spectrum of users' behavior.Firstly,we introduced the end host behavior classification process based on the characteristics of the network connection in detail,and then we introduced the matrix of similarity that is used in our method, and how to structure the connection relation matrix and the connection diagram of nodes based on matrix similarity.Then,we looked at the users' network behavior and the stability of preference trend.Finally,we defined reasonably the nodes' information and the edge weights,using community partition algorithm which called the GN algorithm to acquire the processing nodes connection diagram, the community of end, and had carried on the detailed analysis and explanation.Experiments show that defining the connection relations reasonably based on users' behavior spectrum,and using the community partition algorithm can effectively distinguish between the study of the small network user's behavior to the host model, and carries on the reasonable and meaningful classification.