An Intrusion Detection System Based on KDD-99 Data using Data Mining Techniques and Feature Selection
Pratibha Soni1, Prabhakar Sharma2

1Pratibha Soni, M tech. (CSE), Raipur Institute of Technology, Raipur (CG), India.
2Prabhakar Sharma, Asst. Prof. Department of CSE, Raipur Institute of Technology, Raipur (CG), India.
Manuscript received on June 25, 2014. | Revised Manuscript received on July 03, 2014. | Manuscript published on July 05, 2014. | PP: 112-118  | Volume-4, Issue-3, July 2014. | Retrieval Number: C2316074314/2012©BEIESP
Open Access | Ethics and Policies | Cite
© The Authors. Published By: Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)

Abstract: Internet and internet users are increasing day by day. Also due to rapid development of internet technology, security is becoming big issue. Intruders are monitoring computer network continuously for attacks. A sophisticated firewall with efficient intrusion detection system (IDS) is required to prevent computer network from attacks. A comprehensive study of literatures proves that data mining techniques are more powerful technique to develop IDS as a classifier. Performance of classifier is a crucial issue in terms of its efficiency, also number of feature to be scanned by the IDS should also be optimized. In this paper two techniques C5.0 and artificial neural network (ANN) are utilized with feature selection. Feature selection techniques will discard some irrelevant features while C5.0 and ANN acts as a classifier to classify the data in either normal type or one of the five types of attack.KDD99 data set is used to train and test the models ,C5.0 model with numbers of features is producing better results with all most 100% accuracy. Performances were also verified in terms of data partition size.
Keywords: Decision tree, Feature Selection, Intrusion Detection System, Partition Size, Performance measures.