Integration of heterogeneous classifiers for intrusion detection

Abstract

To address the problem of less rare data and low detection accuracy, The paper proposes a heterogeneous classifier integrated by the random forests, support vector machines, clustering and Bayesian classifier to increase the detecting accuracy of rare class, and to detect rare class with the greatest weighted voting. Experimental results show that utilizing integration of heterogeneous classifiers in intrusion detection system can improve obviously detection precision and decrease false positive rate.