A machine learning-based ensemble model for securing the IoT network

The rapidly growing number of Internet of Things (IoT) devices has led to a rise in data transfers, which has raised security concerns. Due to the devices' limited processing capabilities and vulnerability to many cyber attacks, securing IoT communications is challenging. Security threats, especially Distributed Denial of Service (DDoS) attacks, take a toll on the network in the form of increased communication overhead. Hence, a centralized unit is required to detect DDoS attacks in IoT networks at the earliest. Software-Defined Networking (SDN) promises a potential solution for better network traffic management and data flow. This paper presents a machine learning-based ensemble model for the detection of DDoS attacks in IoT networks using SDN. The proposed model employs a multi-step approach utilizing various Machine Learning (ML) algorithms. The proposed Ensemble Model (EM) combines Logistic Regression (LR), k-Nearest Neighbors (KNN), Gradient Boosting (GB), Extra-tree (ET), AdaBoost, and XGBoost, with XGBoost as the final estimator classifier. Various metrics, including sensitivity, specificity, precision, accuracy, and others, derived from the confusion matrix, evaluate the proposed model's performance. The EM demonstrates superior performance during comparative analysis with state-of-the-art schemes, with a classification accuracy of 99.8%. Furthermore, the paper evaluates the model based on Receiver Operator Characteristic (ROC) curves, showing its superiority in True Positive Rates (TPR) compared to False Positive Rates (FPR). The AUC analysis supports the EM's effectiveness. Cross-validation results further validate the model's robustness, with a mean accuracy of 97.92%.