Deep Learning Models Comparison in binary context for DDoS Attack Detection in Software-Defined Network
Ameur Salem Zaidoun, Zied Lachiri · 2024
Software-defined Networks (SDN) are continually evolving, but they also face numerous security challenges, particularly from Denial of Service (DoS) attacks and their more distributed variant, Distributed Denial of Service (DDoS) attacks. This paper will be focused on showcasing and giving an efficient comparison and classification of Artificial Intelligence (AI) based solutions, mainly Deep Learning (DL) tools, by the experimentation of a set of them on one selected dataset to pick up the most efficient demarcation tool. The main goal here is to separate normal from abnormal traffic without any distinction of different identified attack types using binary classification. Simple and hybrid models will be deployed in supervised mode on a chosen dataset, which is CIC-DDoS2019, that has been optimized for more efficient training. The accuracy results recorded in this case reached ${9 9. 8 \%}$.