Research Of The IP Traffic Identification Techniques Base On Clustering Algorithm

In recent years, with the quick development of the Internet, the network scale extends sharply and network bandwidth increases unceasingly, a wide variety of network applications are emerging. In the same time, a lot of problems are appeared, the expansion of network bandwidth is not able to satisfy the increasing demand of different kinds of network applications, network resources can’t be fully utilized, network attacks and network virus are updated frequently. Network traffic analysis is a powerful tool to be able to effectively solve these kinds of problems, with the help of network traffic analysis tool, network bandwidth can be fully made use of, network security environment can be maintained, and differentiation of QoS control and analysis of network applications can be achieved.Research of the IP traffic identification techniques base on clustering algorithm is a hot spot in network traffic analysis. Clustering algorithms applied to real-time IP traffic identification technology perform very well on IP traffic identification and guarantee higher classification accuracy.In this paper, a deep research was done about IP traffic identification techniques base on clustering algorithms. The main content of the research mainly has the following aspects:1. A simple introduction of recent development of the Internet was done, mainly including the scale of the Internet, the applications of the Internet, the resources and the security of the Internet, meanwhile, purposes and main problems of network traffic analysis currently facing were introduced.2. Current research work of network traffic analysis was summarized. Firstly, three main IP traffic analysis methods were introduced. Secondly, a deep research was done about IP traffic identification based on machine learning algorithms, and four aspects of performance were summarized. 3. Related background knowledge was described, including network data flow knowledge, machine learning basic knowledge, feature extraction algorithm and clustering algorithm.4. Feature extraction plays an important role in IP traffic identification, according to the characteristics of different features in the feature set, features are divided into three groups, meanwhile, normalization is applied in each group in order to decrease the differences between features while increases accuracy scores, and a new feature extraction model was designed.5. Clustering algorithm performance directly determines the IP traffic identification accuracy. On the basis of deep study on a variety of clustering algorithms, DBSCAN algorithm and BIRCH algorithm are used to generate new clustering algorithm which could get better results, and a new IP traffic identification model was designed.