Intrusion Detection Using Isomap and Support Vector Machine

Intrusion detection is still a crucial issue for network security. Support vector machine (SVM) has been successfully applied in intrusion detection systems. However, for further improvement in performance, data dimension reduction should have drawn special attention. This paper proposes a scheme using popular non-linear dimension reduction tool Isomap and one-class support vector machine to detect U2R (user to root) and R2L (remote to local) intrusions. Experiment results on KDDCUP 99 datasets show that our scheme achieves high detection rate for R2L or U2R intrusions and significantly low false positive rate compared with one class SVM alone. It is justified that data dimension reduction is a worthwhile preprocessing stage for achieving high performance in the intrusion detection system.