A new DDoS detection model using multiple SVMs and TRA

Abstract

Recently, many attack detection methods adopts machine learning algorithm to improve attack detection accuracy and automatically react to the attacks. However, the previous mechanisms based on machine learning have some disadvantages such as high false positive rate and computing overhead. In this paper, we propose a new DDoS detection model based on multiple SVMs (Support Vector Machine) in order to reduce the false positive rate. We employ TRA (Traffic Rate Analysis) to analyze the characteristics of network traffic for DDoS attacks. Experimental results show that the proposed model is a highly useful classifier for detecting DDoS attacks. © IFIP International Federation for Information Processing 2005.