Indian Journal of Science and Technology

Abstract

An intrusion detection has a key role in network security that classifies the system activities as normal or suspicious (Anomaly). An intrusion detection system must consistently detect malicious activities in a network and must perform efficiently to manage with the large amount of network traffic. The main objective of this paper is to analysis two issues such as accuracy and efficiency of the system by a novel method of incorporating swarm intelligence with data mining algorithm for feature reduction. The accuracy of the system then can be achieved by several soft computing techniques as Sugeno fuzzy inference system and simplified swarm optimization. The high efficiency can be achieved by Multi-tier approach. The proposed system uses Multi-tier-sugeno fuzzy inference system for fuzzy rule generation that effectively identifying the intrusion activities within a network, Finally, to obtain best result Simplified swarm optimization algorithm are used to optimize the structure of the fuzzy decision engine. The experimentation and evaluation of the proposed method were performed on NSL KDD intrusion detection dataset that shows best accuracy and efficiency than other methods and can easily detect whether the network data are normal or under attack.

Keywords: Intrusion detection, Multi-Tier approach, NSL-KDD, Random Forest, Simplified Swarm Optimization, Sugeno Fuzzy inference system