Detection of DDoS attacks using supervised ML and deep learning approaches for SDN

Software Defined Networking (SDN) has gained attention and has become very popular due to its capability to provide simple and efficient techniques for controlling computer networks. This paper examines how supervised machine learning, unsupervised machine learning and supervised neural network approaches are implemented to detect DDoS attacks. To lessen and avoid DDoS attacks, the analysis procedure requires training and implementing the right model for the assigned network. It has been observed from our extensive evaluation of these learning techniques that K-Nearest Neighbour (KNN) reflects the highest degree of accuracy of 97.50% among all supervised machine learning approaches. In contrast, CNN reflects the highest degree of accuracy of 99.41% among supervised neural network approaches. In addition, we have also evaluated the silhouette score for unsupervised machine learning techniques where K-means with Principal Component Analysis (PCA) outperforms K-means with T-distributed Stochastic Neighbour Embedding (TSNE).