A feature dimension reduction technology for predicting DDoS intrusion behavior in multimedia internet of things

Yongsheng Zong, Guoyan Huang

Abstract: Due to massive data flow and complexity of changeable data characteristics, the dimension disasters problems is ordinary existed in the prediction of the Distributed Denial of Services (DDoS) large-flow attack for multimedia Internet of Things, which will result in the following deficiency such as excessive consumption of computing, and storage resources, reduced analysis efficiency. In this paper, a novel method with the combination of matrix diversity and principal component analysis is proposed for DDoS feature reduction. Firstly, matrix diversity is used to reduce the multiple feature properties of DDoS, and then principal component analysis is used to reduce these features further. Then, the statistical characteristics of these correlations are analyzed. Finally, real-time attack detection is carried out based on mahalanobis distance (MD). It is obvious demonstrated that the proposed method has higher prediction accuracy and more computational efficiency than the traditional method.