SECURING NETWORKS USING SOFTWARE DEFINED NETWORKS AND MACHINE LEARNING

Abstract

Machine Learning techniques are used in Networks to detect DoS and DDoS attacks and to resolves network security issues. As many researchers done their research either on real time datasets or synthetic datasets on different attacks however in our thesis, we aimed to check the performance of Machine learning algorithms, that which one is giving high accuracy in detection of DDoS attack. For this purpose, we have generated simulated datasets in Mininet and in Packet Sender tool. In addition, two well-known datasets has been chosen in which one of them is real time dataset that is ToN-IoT whereas the other one is synthetic dataset Mendeley DDoS. By applying Machine Learning techniques on these datasets, we investigate seven different algorithms: K-Nearest Neighbor, Decision Tree, Random Forest, Logistic Regression, Multi-Layer Perception, XG-Boost, Support Vector Machine and Ensemble Method, results are produced on the basis of accuracy rate. Results are computed on the basis of best features present in all datasets.