Research Of Network Security Audit Based On Data Mining

With the spread of applications of network, matters of network and information security are more and more serious. By analyzing audit data, network security audit technology can detect the current network security state, so, it has great significance in ensuring the security of network information. Data mining has great preponderance in solving current problems of network security audit.Special researches and attentions on the method of network security audit based on clustering analysis were paid on the basis of Data mining and network security audit technology. The main work was described as follows:(1) A method of network security audit was proposed based on clustering analysis, which had improved the general efficiency of clustering analysis by adopting method of feature subset selection aimed to clustering analysis and improved algorithm.(2) Feature subset selection of audit data was researched. Aimed to the character of clustering analysis, a mix-feature subset selection method was proposed. The method could be described as: removing the classifying irrelevant feature to clustering analysis on the base of ReliefF algorithm, then getting rid of redundant feature among the remanent feature by taking advantage of mutual information theory. It had improved the quality of clustering analysis input data.(3) A method of audit data analysis based on fuzzy dynamic was proposed. In method, to fuzzy k-harmonic mean (FKHM) algorithm , a improved algorithm—RFKHM algorithm was proposed, which had reduced the influence of outlier to clustering and improved audit data analysis analysis efficienc.(4) On the basis of reaserchs of feature subset selection of audit data and audit data analysis based on fuzzy dynamic clustering, a network security audit system of double-data engine was designed.