Research On Discretization Technology And Its Application On Intrusion Detection

With quick development and application of computer and network technology, Data mining that comes into existence as the situation requires is greatly concerned. Meanwhile, Intrusion detection shows its strong vitality. Association rule is such an important and active technique that it can discover the behavior pattern between normal and abnormal data. Thus, research of association rule is in favor of improving the accuracy and timeliness. Generally, according to current set theory in math, the large real numbers need to be discretized so that this study has important theoretic and practical significance.For the discretization, the complementarities and relativities between different attributes are not taken into account in most discrete methods. Every processing tends to perform relatively so that it always changes the equivalent relation and generates unreasonable and redundant breakpoints. Among ID data, continuous attributes are so abundant that single attribute discretized inefficiently and the results with many break points often lead to make irrational and redundant.This thesis presents an algorithm with the result that the breakpoints are not only in line with actual statistical distribution, but also they are more reasonable and refined. The mult-attribute algorithm shows the reduced redundancy, which compared with related studies its rationality and accuracy has been improved. Refering to the aspects of how to apply the discretization method into intrusion detection data and how to discretization assists associative technique to be adopted into intrusion detection, the results of research has its development potential and effectiveness.