Network Application Identification Based On Wavelet Transform And Adaptive Resonance Theory Algorithm

The network develops rapidly, accurate identification and categorization of network traffic is more and more important.The identification of network application based on port number is constrained by the status that many peer-to-peer applications make use of dynamic port numbers. And considerations of safety, applies enciphered data, traditional methods based on payload and well-known port numbers have not held good.The method based on application layer statistics is used to cope with this problem.Machine learning is the essence of the method based on application layer statistics; the K-means clustering is one of methods to solve this problem. It deals with the non-prelabeled data in the classification problems, but it is time-consuming and low accuracy. In this paper, two improvements based on K-means are proposed. One is doing data preprocessing with wavelet transformation prior to K-means algorithm. The time for data processing is shortened; the other doing Adaptive Resonance Theory prior to K-means algorithm to improve accuracy. Experiments in Campus network have corroborated the strong points, the storage of data has been reduced; and finally, the classification precision of this method is increased.