Mitigation of DDoS attack using Machine Learning Algorithms in SDN_IoT environment

Abstract

DDoS is one of the major problems in security aspects. Denying a site, application or server from performing their desired operations is usually called DDoS. This type of attack is focused mainly to tamper with the resources and make them unavailable. Making the service unavailable is usually achieved by manipulating network packets. This idea uses Machine Learning Techniques to eliminate and mitigate DDoS and provide real-time protection for vulnerable resources. The usage of a trained Machine Learning Model can be able to detect DDoS occurring through various manipulating protocols like TCP, UDP and ICMP. The dataset is cleansed to get rid of unknown values. Then out of all the available features in the dataset, the most important attributes for DDoS occurrence is selected and targeted as the main features. The training is done based on the DDoS occurrence attributes available versus the target attribute. The target attribute here is the result which is nothing but both the cases of DDoS occurring. Random Forest Classifier model is used along with four other classification techniques which are K-Nearest Neighbour, Multi-Layer Perceptron classifier, Decision Tree Classifier and Logistic Regression. The usage of various algorithms is to select the most suitable algorithm for a particular protocol. After training the model, the model is tested for various algorithms. The Decision Tree algorithm has performed well for the cases of attack with UDP protocol and TCP protocol. KNN has outperformed other algorithms in the cases of attack with ICMP protocol.