Study of the Effectiveness of ML Algorithm Parameters in Classifying DDoS Attack in SDN Networks

Abstract

Software Defined Network (SDN) is the new era of networking because of the advantages it offers. But They suffer from multiple security threats and attacks targeting their vulnerabilities. Perhaps the most prominent of these attacks are Distributed Denial-of-Service (DDoS) attacks. Machine learning techniques are increasingly used in predicting security attacks. In this paper, we will present a practical study of a set of machine learning algorithms for predicting DDoS attacks.

The study based on a set of algorithms, and for each algorithm, we defined the best and worst parameters comparing on the basis of the accuracy, f1-score scales and whether the algorithm was appropriate for real-time applications. The results showed that the decision tree algorithm performed the best with an accuracy of 99.99%, while the Multinomial NB algorithm performed the worst with an accuracy of 64.36%. The SVM algorithm had the longest training time at around 76 seconds, while the decision tree algorithm had the best time at 0.018 seconds. In terms of F1 score, the decision tree algorithm was the best at 99.99%, while the worst was the Multinomial NB algorithm at 69.26%.