Software Defined-Network Intrusion Detection Model Using Stacked Ensemble Techniques of Machine Learning

Abstract

Software defined-network (SDN) brought in so much of flexibility in network management and administrations through its programmability and centralized nature. However, this programmability, exposes SDN to constant evolving network attacks. To address this challenge, previous studies have shown that intrusion detection system (IDS) is very effective. So many approaches were adopted to develop IDSs especially machine learning because of its strength in detecting trends in a given data. Unfortunately, this strength depends greatly on the quality of the training dataset which is subject depreciation over time. Couples with the constant evolutions of network attack, the depreciations in quality of IDS training datasets have made it very difficult for machine learning IDSs to detect attacks
accurately. In order to address this challenge, this study proposes a software
defined-network-based intrusion detection model using stacked ensemble
technique of machine learning. The study adopts inSDN dataset as the training
dataset because of its of quality in SDN features. From the experimented result, the
model performed very well by recording 99.3% of accuracy. Despite the
performance of this model, the model has never been evaluated in a real SDN
environment.