The Method Of Traffic Identification Method Based On Clustering Algorithms And System Implement

Precise and quickly identify traffic is the basic of the guarantee of network security and traffic control. Especially in today, Internet technologies are developing rapidly. Many new protocols and applications continually emerge, such as P2P (Peer-to-Peer) etc.. The continually increasing new transactions resulted in severe overhead in bandwidth, even causing network congestion, especially in enterprise or campus network. Meanwhile, downloading applications from an unsafe network may give chances for virus and malicious code to invade the system. The main tasks of this paper are,1. This paper has studied the network transport attributes of applications, and several clustering algorithms which are oriented to traffic identification. This paper has analyzed the advantages and disadvantages of each attribute and clustering algorithms, determining direction for the research aim of this paper.2. In the phase of designing traffic identification algorithm, firstly, this paper chooses the size and direction of the first four packets of applications which are connecting to network to be the identification index. Secondly, take K-means algorithm for instance, because there are some defects in choosing initial centers, our algorithm improves this process and introduces NMI to validate the performance of clustering, constructing an improved K-means algorithm which is used in clustering and abstracting characteristics.3. After gaining characteristics, aiming at the frequently changing characteristics of network to modify characteristics parameters in code would course problem of compile time too long, reducing system performance. This paper proposes using XML script language to describe traffic characteristics, and programs match traffic via loading these characteristic-describing files. When the characteristics have changed, it just needs to modify characteristics parameters in file.4. Based on above three points, this paper designs and implements a traffic identification system based on clustering algorithm. This system supports a variety of clustering algorithms, and uses XML script language to describe traffic characteristics, identifying traffic online by loading characteristics description file.