DETECTION OF DDoS ATTACKS IN CLOUD ENVIRONMENT USING COLLABORATIVE TRAINING

Abstract

Distributed Denial of Service (DDoS) attacks are serious threat to Cloud. These attacks consume large amount of resources and increase the service usage cost by significant factor. Due to multitenancy and self-provisioning properties of Cloud, traditional DDoS detection techniques cannot be directly applied to it. Hence there is a need of Cloud specific DDoS detection framework. In this work, we have proposed a behavior based framework uses Euclidean Distance for finding the hidden correlation between different attributes of the same network traffic record. After extracting the hidden correlations, Mahalanobis Distance is used for finding correlation between them. A collaborative training and warning sharing mechanism has also been proposed for adopting any changes of network behavior in Cloud and for preventing the occurrence of same types of attack in future. Performance improvement on threshold point provides extra flexibility to the network administrator for selecting an efficient threshold point. Our system performance evaluation shows very high detection rate with few false positive and false negative rate.