A new approach of extremely randomized trees for attacks detection in software defined network

Abstract

Software defined networking (SDN) is the networking model which has completely changed the network through attempting to make devices of network programmable. SDN enables network engineers to manage networks more quickly, control networks from a centralized location, detect abnormal traffic, and distinguish link failures in efficient way. Aside from the flexibility introduced by SDN, also it is prone to attacks like distributed denial of service attacks (DDoS), that could bring the entire network to a halt. To reduce this threat, the paper introduces machine learning model to distinguish legitimate traffic from DDoS traffic. After preprocessing phase to dataset, the traffic is classified into one of the classes. We achieved an accuracy score of 99.95% by employing an optimized extremely randomized trees (ERT) classifier, as described in the paper. As a result, the goal of traffic flow classification using machine learning techniques was achieved.