An efficient LGBM based DDoS attack Detection Approach for SD-IoT

Internet of Things(IoT) applications have grown more prevalent, making security a top priority. Many IoT devices participated to a recent distributed denial-of-service (DDoS) attack without their owners' knowledge. To control IoT devices securely, the new SDx paradigm is evolving. This study begin by outlining a high-level architecture for the SD-IoT, which is based on the SDx paradigm. The objective of this work is to efficiently detect the DDoS attack in the controller of the SD-IoT. For this, a dataset created in the control plane of SDN has been used for training some well-known classifiers namely LGBM, SVM, Random Forest, and KNN, with the objective to find the best performing classifier under some performance measuring metrics namely; accuracy, F1, recall, precision, Cohen's Kappa Coefficient, False Alarm Rate(FAR) and execution time. It is found that LGBM is outperforming all the classifiers as it is giving values more than 99.72% values for accuracy, precision, recall, and F1, and 0.35% for FAR, 4.134 second for execution time, and 99.42 for CKC. Finally, the performance of this best performing classifier i.e., LGBM is compared with some similar state-of-the-art works. It is found that again it is outperforming other works.