期刊名称：Faculty of Computer Science and Information Technology
出版年度：2008
卷号：0
期号：0
语种：English
出版社：Faculty of Computer Science and Information Technology
摘要：Attacks prediction is the action needed by an intrusion detection system as a firststep or in case of an attack. Many methods can be done to predict types ofattacks. One method used is the technique of data mining. But not all datamining algorithms have good performance in classifying the type of attack.Therefore, this study will try to compare several algorithms. There are 41attributes / variables that are used to classify types of attacks. Of the many typesof attacks that happened, then grouped into four classes, which are categorizedbased on the final goal is achieved by an attack. Categories include: Probe, DoS,U2R, and R2L. And for the data set used is the data set from KDD Cup 1999,where this set of data is referenced data for IDS case study. The comparisonalgorithm will be seen based on the value of instances classified correctly,incorrectly classified, kappa statistic, true positive, false positive, and theconfusion matrix. In this research, the comparison is done by comparing SVMalgorithm, decision tree and Naive Bayes. By using the tool Weka (WaikatoEnvironment for Knowledge Analysis) version 3.4.13, it can be concluded thatthe algorithm has a superior performance is a decision tree.
