A Method of Abnormal Behavior Classification in Mobile Internet

With the large-scale employment of mobile internet, more mobile abnormal behaviors appear and are a great threat to the security of mobile users. Since many of those are unprecedented and unknown, it is difficult to classify these abnormal behaviors from normal behaviors by use of existing security application. In this article, a new method is proposed to classify and detect the abnormal behaviors in mobile internet. Firstly three classifiers are designed to divide all of the normal and abnormal mobile internet behaviors based on the characteristics of these behaviors. These three network behavior classifiers are trained by the known abnormal behaviors and normal behaviors data of mobile internet respectively. The algorithm of these classifiers is an improved multi-category Naïve Bayesian algorithm. Those classifiers analyze the corresponding type of network behaviors to detect the new abnormal behaviors. A self-learning procedure is adopted to gradually optimize those classifiers for different behavior. A mobile abnormal behaviors analysis system based on the above behavior classifiers is described.