A Centralized LGBM based Attack Detection and Mitigation Approach for Software Defined Networks

Software Defined Networking is one of the most emerging technologies in the field of network architecture due to Its capability to centrally observe all the network-related activities happening in the network. On one side, where SDN is getting popularity, the attacks are consistently increasing day by day. In the current work, a centralized approach to detect and mitigate the attack in the controller of SDN has been proposed. For this, a dataset created in the controller of SDN has been used to train some well-known tree-based classifiers namely XGB, RF, ET, LGBM, and GB. They have been evaluated under 8 metrics, namely execution time, recall, FAR, accuracy, F1, precision, CKC, and AUC-ROC where LGBM is found outperforming other classifiers under all the metrics, so it deployed in the controller to detect and mitigate the attacks. The values given by LGBM for accuracy, precision, recall, and f1 are more than 99%. For FAR, CKC, and AUC its value is 0.5%, 98.54%, and 0.995, respectively.