Abstract:Most detecting algorithms which were confined to analyze packet could not describe DDoS completely. Without complete description

detection would be effected

thus could not provide effective information for network management. Concerning this issue

this paper proposed a novel approach based on packet-and-conversation-considering-with-context(PCC) time series analysis to detect DDoS attacks. By analyzing packet and conversation of DDoS traffic simultaneously

and combined with context information

multi-dimensional characteristics can describe DDoS better. And then support vector machine (SVM) classifier was adopted to establish the DDoS attack detection model from packet feature vector and conversation feature vector. Furthermore

a reliable alarm strategy was proposed to further reduce the influence of noise. The experiment results showed that PCC based approach can detect DDoS attacks effectively and reduce the influence of traffic noise.