Detecção de DDoS Através da Análise da Quantificação da Recorrência Baseada na Extração de Características Dinâmicas e Clusterização Adaptativa

Abstract

The high number of Distributed Denial of Service (DDoS) attacks have demanded innovative solutions to guarantee reliability and availability of internet services. In this sense, different methods have been used to analyze network traffic for denial of service attacks, such as neural networks, decision trees, principal component analysis and others. However, few of them explore dynamic features to classify network traffic. This article proposes a new method, called DDoSbyAQR,that uses the recurrence quantification analysis based on the extraction of dynamic characteristics and an adaptive clustering algorithm (A-kmeans) to perform better classification of the attack network traffic. The experiments were done using the CAIDA and UCLA databases and have demonstrated ability to increase the accuracy (98.41%) of DDoS detection.