Aode Intrusion Detection Research, Based On Attributes Reduction

With the rapid development of computer technology, information networks have become an important guarantee for social development. However, any form of network services will lead to security risks. Network security is a complex project, purely passive defense is ineffective. Intrusion detection system as the second safety barrier after the firewall, using of active defense technology, has become an important research topic in computer network security. In my paper, the Bayesian method has been used to research the intrusion detection.For the first, the paper introduces the Bayesian theory, and then uses the simplest naive Bayesian classifier builds intrusion detection model; and on the weka platform, validated by experiments naive Bayesian classifier built intrusion detection model is simple and effective. But the naive Bayesian classifier assumes that the condition attributes are mutually independent, it is difficult to meet the actual situation, which leads to its classification accuracy decreased. Therefore, the paper uses AODE classifier, which weakens the Naive Bayesian classifier assumes conditional independence. On the weak platform through experiment, AODE has been proved has the higher classification accuracy than the naive Bayesian classifier.In the intrusion detection system, the data pre-processing is also very important. Whether used the naive Bayesian classifier or the AODE classifier to build intrusion detection model, do not take into account the redundancy of data attributes; So in the paper we use the theory of rough set to reduce attributes, removal of redundant attributes. Finally, we through experiments to verify the Naive Bayes classifier and the AODE classifier are more effective after the attribute reduction.