Distributed Intrusion Detection Method Based
on the Diversity of Immunity
Gu Yu1£¬2£¬Zhao Jiashu3£¬Zhang Tianjun2£¬Xu Zongben3
£¨1.School of Electronics and Information Engineering£¬ Xi'an Jiaotong University£¬
Xi'an 710049£¬China£»
2.Network Information Centre£¬Yunnan Nationalities University£¬ Kunming 650031£¬ China£»
3.School of Sciences£¬ Xi'an Jiaotong University£¬Xi'an 710049£¬China£©
Abstract£ºBased on the diversity of immunity£¬ a distributed intrusion detection method
is presented£¬ in which the support vector machine£¨SVM£© is used as an antibody to
detect intrusion. In particular£¬ the random subspace method is introduced to create
various SVM individuals£¬ and then the individuals are evolved with artificial immunity
algorithm. By introducing Qª²statistical and mutual information as the measures of
the antibody¡¯s diversity£¬ the detector group with the characteristic of mutualª²supplied
knowledge can be obtained. At last the results of all detectors in the population are
integrated by the ensemble idea. This distributed parallel detection can strengthen the
robustness of the systemª± Experiments show that this method can produce antibodies with
better diversity£¬ and the detection accuracy is better than a single SVM and Bagging
methodª±
Keywords£ºintrusion detection£» artificial immune£» diversity£» support vector machine