Principal component analysis of network security data based on projection pursuit

Abstract

In network security situation awareness system, the data are characterized by huge quantities, numerous features, redundancy, etc. These features may seriously impact the efficiency of situation evaluation and prediction. This paper proposes a principal component analysis algorithm based on projection pursuit (PP-PCA) to solve these problems. Combined with particle swarm optimization and exterior point penalty function, PP-PCA projects the data onto one-dimensional plane then figures out several composite indicators which play leading roles. The simulate experiment shows that it can overcome the redundancy and improve the efficiency of the situation evaluation and prediction. © Springer-Verlag Berlin Heidelberg 2012.