A Research on the Identification of Internet User Based on Deep Learning

Abstract

In the environment of big data, analyzing internet user behavior has become a research hot spot. By profiling the normal online behavior data of network users to learn their online habits and preferences, is not only helpful to provide network users with more efficient and personalized network services, but also to update the network security policies. Because there is no identification of network users in network management, network administrators need to develop and deliver relevant network services manually to user base on the network user Internet Protocol (IP) address. Therefore, this paper proposes the utilization of deep learning technology to identify network user automatically after fully understand the behavior of network user. At the first, a network identification model based on Deep Belief Network (DBN) is proposed. Then, we apply the Tensorflow framework to construct a DBN model suitable for network user identification. Finally, an experiment with real data set was undertaken upon the model to verify its accuracy on identifying network users. It is found that DBN-based identification model can achieve a high classification accuracy of user identity by constructing deep network structure.

This work was supported by a grant from the Key Research and Development Program of Zhejiang (No. 2017C03058), Zhejiang Provincial Key Laboratory of New Network Standards and Technologies (NNST) (No.2013E10012).