Analysis of KDD CUP Dataset Using Multi-Agent Methodology with Effective Fuzzy Based Intrusion Detection System

Abstract

ABSTRACTSecure environment construction against the several intrusions is an attractive research area in Wireless Sensor Network (WSN) and Mobile Ad-hoc Network (MANET)-based real-time applications. Intrusion Detection System (IDS) evolved in network research studies improves the security and protects the resources from the various intrusions. The centralized and the distributed IDS are the major IDS categories. The central point failure in centralized IDS under heavy load introduces the security issues. The centralized system deficiency and hierarchical nature introduces the agent-based IDS. This article proposes the Multi-Agent IDS to detect the packet transmission failures and anomaly behavior in two different datasets namely KDD cup 1999 and the real-time traffic dataset. The Multi-Agent Based IDS (MAIDS) employs the distance and density-based algorithms for cluster formation. The rules formation in either association or sequential manner detects and classifies the attacks to the respective agent. Finally, the fuzzy-rules formulation in MAIDS predicts the intrusion type. The comparative analysis between the MAIDS and Network Intrusion Detection System (NIDS) regarding latency, overhead, and packet delivery ratio conveys that the MAIDS provided the better performance than the NIDS.