作者: V KATOS
DOI: 10.1016/J.INS.2007.02.034
关键词:
摘要: This paper evaluates the statistical methodologies of cluster analysis, discriminant and Logit analysis used in examination intrusion detection data. The research is based on a sample 1200 random observations for 42 variables KDD-99 database, that contains 'normal' 'bad' connections. results indicate more effective than or detection. Specifically, according to Kappa statistic makes full use all information contained confusion matrix, (K=0.629) has been ranked first, with second (K=0.583), third (K=0.460).