Detecting DDoS Attacks Using Dispersible Traffic Matrix and Weighted Moving Average

Abstract

Distributed Denial of Service (DDoS) attacks have become significant threats on Internet according to the development of network infrastructure and recent communication technology. There are various types of DDoS attacks with different characteristics. These differences have made very difficult to detect such attacks. Furthermore, the sophisticated the evolution of DDoS attacks techniques and the enhanced scale of Botnet encourage attackers to launch DDoS attacks. The IP spoofing technique also makes difficult detect and traceback of DDoS attacks. In this paper, we propose a new detection model for spoofed DDoS attacks using dispersible traffic matrix and weighted moving average. This proposed detection model can not only visualize network traffic streams but also describe the dispersibility characteristics of DDoS attacks such as intensity, duration and rate of DDoS traffic. We carry out experiments on both DARPA 2000 dataset and real data in our network testbed environments so as to validate the feasibility of our approach. Our approach demonstrates that it effectively detects the DDoS attacks in the early stage and in very short time, even though DDoS attacks' streams are low. Also, the proposed detection model shows a good performance in terms of detection accuracy, speed, and false alarms.