一种网络数据包的有效分类方法  

AN EFFICIENT CLASSIFICATION METHOD FOR NETWORK DATA PACKET

作　　者：阳万安[1] 胡其华[2] 

机构地区：[1]宜宾学院计算机科学与技术系,四川宜宾644000 [2]宜宾学院数学与应用数学系,四川宜宾644000

出　　处：《计算机应用与软件》2010年第2期118-120,129,共4页Computer Applications and Software

基　　金：四川省教育厅(2006C047)资助

摘　　要：终端用户的各种网络行为都会产生大量的数据包,在没有其它任何先验知识的情况下,如何把数据包分成网络游戏包和非网络游戏包是一件困难的工作。通过对大量的数据包样本进行分析,发现用地址偶对和端口偶对的线性拟合斜率作为模式的两个特征,再构造出一个线性分类器,采用增量校正算法求解该线性分类器的权向量,进一步利用该线性分类器,可以对各种环境下新的数据包准确划分为网络游戏包和非网络游戏包,据此可以控制终端用户网络游戏行为。Various network actions of the end users will produce a large quantity of data packets. How to sort out the data packets as the game or non-game packets of the networks without any priori knowledge is a difficult work. After analysing quite a few packet samples, we found that to use the linear fitting slopes of the address pair and the port pair as two features of the pattern, and then to construct a linear classifier, followed by to obtain weight vector of it with increment revision algorithm and to further use the classifier, in this way it is able to partition the new packets in various conditions into the game packets or non - game packets accurately, according to this the end users＇network game actions can be controlled .

关 键 词：先验知识 地址偶对 端口偶对 模式 分类器 

分 类 号：TP393.08[自动化与计算机技术—计算机应用技术] TP391.4[自动化与计算机技术—计算机科学与技术]