A Method for Analyzing Software Faults Based on Mining Outliers' Feature Attribute Sets

摘要

Faults analysis is a hot topic in the field software security. In this paper, the concepts of the improved Euclidian distance and the feature attribute set are defined. A novel algorithm MOFASIED for mining outliers' feature attribute set based on improved Euclidian distance is presented. The high dimensional space is divided into some subspaces. The outlier set is obtained by using the definition of the improved Euclidian distance in each subspace. Moreover, the corresponding feature attribute sets of the outliers are gained. The outliers are formalized by the attribute sets. According to the idea of the anomaly-based intrusion detection research, a software faults analysis method SFAMOFAS based on mining outliers' feature attribute set is proposed. The outliers' feature attributes can be mined to guide the software faults feature. Experimental results show that MOFASIED is better than the distance-based outlier mining algorithm in performance test and time cost.