The Cluster Model On Terminal Behavior Of Trusted Networks Based On GEP-CPN

The network security has become increasingly manifest its vulnerability in current network development. For a growing number of network attacks, the relevant researchers or technical staffs propose appropriate and effective remedial measures. For example, a variety of firewall, antivirus software and the security software for a specifically field, etc. But, these protective measures is designed to prevent external attacks as the primary means, the security problem of the network terminal itself did not be taken into. On the security issues, the insecurity and incredible of the terminal is the key reason in the existence of network security threats. Because of this, domestic and foreign scholars have presented the concept of the trusted network in recent years, trying to make a feasible measure for these issues, at the same time, these scholars get some research and application results.In the trusted network, especially the user terminal behavior is a major research focus. For example, how make the properties of the terminal behavior as a unified formal description. In the terminal connecting to the network, how to build the terminal authentication mechanisms, how to solve a user terminal submit its request to the service provider for identifying and determining the credibility level of their behavior, and so on.Based on these, this paper present an improving method based on neural network for predicting and control the terminal behavior of the Trusted Network, to achieve better protection of network security and credibility. First, study the current problems in the trusted network, and the key thechnique that how to extract the evidence in credible field. And then research the general assessment method of the terminal users'evidence. Second, combined with the counter propagation network (CPN) mechanism and gene expression programming (GEP) algorithm, designed the model of clustering on the terminal behavior of trusted networks based on GEP-CPN, Finally, the validity and superiority of the presented algorithm is demonstrated by experiments.This paper makes mainly the following contributions:a) Analyse the definite and characteristic of the terminal user evidence in trusted network, obtain and standardize the evidence. And then obtain data set by using these metods, that is: the various relevant characteristics in the evidence of the different users'behaviors are studied.b) Study how to improve the CPN network for jumping off the shakles of the traditional CPN network, to achieve best training networks. That is: optimize the input vector and enable the weight vector of these neurons in the Kohonen of the traditional CPN network to consistent with the actual input vectors as possible as it can.c) Evaluated the clustering result on the new model, and compared the various cluster methods with the same text data set.d) Before making the right judgement for the network itself, the evidecnces of the terminal users'behaviors in network must be anlysised. And cluster analysis can extract the useful links between the different data in the data set. So this paper cluster and analy the data of behavior in the network, experiments show that the model have some improvements on the security of the network, and can effectively provide consultation for protecting the entire network.