A Data Mining Approach for Intrusion Detection System Using Wrapper Based Feature Selection Method

No Thumbnail Available



Journal Title

Journal ISSN

Volume Title


Addis Ababa University


Data mining techniques can be used for network intrusion detection systems. Network traffic data are usually large in number of instances and in number of features. Different techniques are available in reducing large data set size. Applying data mining techniques on a large data set may result in wrong output and may also imply computational and time cost. Feature reduction by reducing redundant and irrelevant feature may be a good approach in finding optimal data for applying classification algorithms and finding out accurate result. The main challenge in building intrusion detection system is building a system which can identify newly introduced attack types which were not included in training set. Most intrusion detection systems are developed meant to identify already trained intrusion or attack types. The objective of this research is to explore the possibility of developing a predictive model for intrusion detection using efficient wrapper based feature selection technique. The data used for the research is NSL KDD data set which is a network traffic data with manually injected network intrusion attempts. In this research a wrapper based feature selection approach is used to identify an optimal subset of features from NSL-KDD data set. After applying wrapper based feature selection and using the induction algorithm to analyze the KDD data set a promising result has been obtained in classifying the different attack types in the NSL- KDD data set. The test set which has new attack types that were not included in the training data set seem to be effectively classified using the classification models built. Using the predictive model built, the attack types that were correctly identified were 95.16%. Which was a better result compared to the same algorithm being applied on data set on which filter based feature selection is used. Since wrapper based feature selection uses classification algorithms in evaluating the relevance and optimality of a feature the time complexity is significant. Finding techniques for reducing this problem is important by using techniques like reducing stopping criterions or by experimenting different combination of searching and evaluating algorithms.



System Using Wrapper Based Feature Selection