Forward Feature Selection for DDoS Detection on Cross-Plane of Software Defined Network Using Hybrid Deep Learning

Abstract

This paper presents a Forward Feature Selection (FFS) method implemented in selecting the best features from the Canadian Institute of cybersecurity intrusion detection system for distributed denial of service attacks. Deep Learning an advanced machine learning approach was used to implement a hybrid method of combining two deep learning algorithms of Convolutional neural network and Deep neural network. Experimentation was used on feature selection to detect distributed denial of service on the software-defined network using the FFS method. FFS method offered an accuracy rate of 98.09%, specificity 97.93%, false-positive rate 0.0207 %, recall 98.32%, precision of 97.31 % and F1-score 97.81%