The Naive Bayse Classification In Data Mining And Its Application In Intrusion Detection

With the development of society, the progress of human civilization, thedevelopment of Internet is unprecedented, it can be said of today’s human societydevelopment is inseparable from the network. With the development of Internet, thenetwork needs to deal with the amount of information is unimaginable, the networkis like a double-edged sword, brings benefits and convenience to mankind, it alsobrings a lot of negative effects, the network attack and damage is also increasingyear by year. The current network attack has become more complex and intelligent,so traditional security defense techniques (such as firewall, access control and so on)have become difficult to deal with. Intrusion detection system (IDS) as the maintechnical to protect the security of network becomes concern. After years ofdevelopment, the technology has become more mature,. the application of datamining in intrusion detection of network security has become a research hotspot.However, due to the continuous development of the method of intrusion and theincreasing of intrusion information, the traditional data mining technology used inIDS has become stretched, unable to ensure the detection rate and the real-time. Thispaper is based on traditional Bayes classification algorithm, proposed an improvedBayse Classification algorithm and an improved system model, aims at improvingthe traditional Bayes classification of intrusion detection system model in thedetection rate, detection time. After this, put forward a kind of attribute reductionmethod based on the rough set theory, in order to reduce the complexity of theattribute, delete redundant attributes, reducing the time of modeling of the system.The main research work are as follows:(1) Analysis of the data mining technology and intrusion detection technology,analyzed the current development trends of this field, explained the concept andclassification of intrusion detection technology.(2) Analysis of Naive Bayse Classification algorithm, on this basis, proposed animproved Bayse Classification algorithm and improved the traditional intrusiondetection system model, in the improved system model, combined the patternmatching method, although in the system has added a new module, in the early stageof building the system, may increase the workload, but through the improvement tothe traditional model,, can improve the detection efficiency of the system and reducethe test time in the intrusion detection process.(3) The attribute reduction method based on rough set theory are studied andanalyzed, study on the method of the distinguish matrix in rough set theory, and studies the attribute reduction process, pointed out the lack of distinction matrixmethod, on this basis, put forward a attribute reduction method based on dependencedegree, and gives the solving process of attribute reduction, by comparison, provedthat the improved method is more outstanding than the traditional method in the timeand space performance.