Performance Evaluation of Machine Learning Approaches for Classification of Ddos Attacks in Software Defined

Abstract Enhancing network management and centralized network control is the goal of software-defined network. It separates the data plane from the control plane. Pros of SDN include its programmability, flexibility, and manageability to accommodate network users' ever-increasing demands. SDN is known to be a strong and safe defense against malicious network activity, but it is also susceptible to attacks like distributed denial-of-service attacks. This is a result of its centralized control features. To defend against this threat, a number of machine learning strategies have been employed. This study evaluated the performance of five Machine learning (ML) Algorithm (Naive Bayes, K-nearest neighbour, Random Forest, support vector machine, Logistic regression) for classifcation of traffic dataset as either normal or DDOS attacks in SDN. The study adopted 10-fold cross validation on each of the ML Algorithms and the result showed Random forest has the highest performance compared with other machine learning model while Naive Bayes had the lowest training time.