Performance Measuring Tool for Data Mining Techniques in Intrusion Detection System

• 主要作者: Roslan, Muhammad Firdaus
• 格式: Final Year Project
• 语言: English
• 出版: Universiti Teknologi PETRONAS 2006
• 主题: T Technology (General)
• 在线阅读: http://utpedia.utp.edu.my/9309/1/2006%20-%20Performance%20Measuring%20Tool%20for%20Data%20Mining%20Techniques%20in%20Intrusion%20Detection%20System.pdf; http://utpedia.utp.edu.my/9309/

The research project is about to develop a performance measurement tool for Data Mining (OM) techniques in Intrusion Detection System (IDS). Basically, IDS is a network security system that is used to detect cyber attacks intrusion. By applying the Data Mining technique it might improve its accuracy as well as its efficiency in intrusion detection process especially in a large and fast network. However, there are various kinds of techniques in OM that can be used to enhance the intrusion detection process in IDS such as K-mean clustering, Support Vector Machine (SVM), Self Organizing Maps (SOM), Neural Networks, etc. Therefore, a performance measurement is required in order determine the best OM technique to be used depending on the network environment and the type of the IDS used. The performance measurement takes place at the final stage of the Knowledge Data Discovery (KDD) process which a step by step procedure in implementing the DM techniques. With the help of this new tool, it can reduce the human intervention in performance measurement process as much as possible by replacing the manual tasks with the automated approach. As a result, errors due to human conducts can be reduced. This is because a slight of error might affect the overall performance results measured. The final results are so important that it is to be used in decision making of the implementation of OM technique in IDS. The tool is comprised of three main modules: confusion matrix analysis, calculation of the detection rates and the false alarm rates, and generating the ROC curves as the final result.