OCFSII: A New Feature Selection Based on Orthogonal Centroid both Inter-class and Intra-class for Vulnerability Classification

With the rapid development of information technology, vulnerability has become a major threat to network security management. Vulnerability classification plays a vital role in the whole process of vulnerability management. It is the key point to select proper features to represent categories. Due to the low efficiency and accuracy of some common feature selection algorithms, in this paper, we proposed a new method called OCFSII, which measures the importance of the feature terms both in inter-class and intra-class based on orthogonal centroid. We evaluated the method on the vulnerability database, using two classifiers, namely, KNN and SVM. The experimental results show that the proposed method OCFSII outperforms Information Gain (IG), Document Frequency (DF), Orthogonal Centroid (OC), and is comparable with Improved Gini index (IGI) when KNN used while OCFSII is superior to the four algorithms. In addition, OCFSII is more advanced than OC.