期刊名称：International Journal of Computer Science and Information Technologies
电子版ISSN：0975-9646
出版年度：2016
卷号：7
期号：3
页码：1238-1241
出版社：TechScience Publications
摘要：Nowadays a trend is being introduced in which thetraffic over the Internet do not have predefined specificationssuch as port numbers. So they cannot be classified easily froma huge mixed traffic. The main reason behind not havingproper specifications for application protocols is to penetratefirewalls or escape administrative control. Many systems arepresent that can identify and classify the mixed traffic overnetwork but require prior protocol specifications. So thesesystems do not work well for the current emerging trendmentioned above. Such systems find their application inmainly deep packet inspection module of IDS systems whichnowadays are increasingly introduced in large scale in variouscompanies. In this paper, we propose a system that exploitsthe semantic information from the raw network traces andclassify them into identified application protocol withoutrequiring a need for prior protocol specifications. The featureextraction module in the system is being implemented usingCollapsed Variational Bayesian Inference Algorithm as inprevious existing system the same module was implementedusing Gibbs Sampling algorithm which is an iterativealgorithm that is very time consuming. Our aim would be toevaluate the quality of our proposed system by calculatingthree metrics namely recall, precision and F-Measure and thencomparing the results with those of prior existing system thatuses Gibbs Sampling algorithm to find whether our approachis an efficient one or not.
关键词：Network traffic classification; application;protocol; deep packet inspection.
