A Framework for Simulation of Intrusion Detection System using Support Vector Machine

Abstract

intrusion compromises the security and the value of a computer system in network. Legitimate users find it difficult to access network services due to the network attacks as they intentionally occupy or sabotage network resources and services. The intrusion detection system defends the critical computer system and networks from cyber-attacks. Various techniques of machine learning are applied to intrusion detection system. In this paper, a framework for simulation of intrusion detection system is described. The radial basis kernel based support vector machine is used to simulate the intrusion detection system. The major research goal regarding the SVM is to improve the speed in training and testing by determining the best kernel for a given data. Out of the various parameters of the packet only few important normalized parameters are used which will result in improving speed of training the SVM and high detection rate. The KDDCUP'99 dataset is used to train and test the system. The experimental results show that the detection rate of the system is 88.27% with good speed. Furthermore, two applications of framework are described to show how the system can be used to generate pattern of attack for testing the system and how the system prevent downloading of large PDF files from server by unauthorized user.