Advanced Intrusion Detection in Software-Defined Networks through Ensemble Modeling

Abstract

In software-defined networks (SDNs), effective intrusion detection is crucial for maintaining network safety and integrity. Traditional intrusion detection systems (IDS) have often failed to identify sophisticated threats due to their limited detection capabilities. To address this challenge, this study introduces an ensemble model that integrates Convolutional Neural Networks (CNN), Support Vector Machines (SVM), and XGBoost. This ensemble approach aims to enhance intrusion detection in SDNs by leveraging the strengths of each model. Using the InSDN dataset for training, our proposed model demonstrates superior performance and significantly outperforms a set of state-of-the-art models achieving a performance of 95% for accuracy, precision, recall, and F1 score exceeding the performance of other methods. Additionally, it significantly reduces false positive rates, highlighting its effectiveness in detecting complex intrusions in SDNs.