Evaluating the efficiency of Supervised Learning Classification Models in Solving Denial of Service Attacks Problem

Abstract

Machine Learning has been widely used in several disciplines nowadays. The rabid development of Denial of service attacks outdated traditional methods for network security regarding the DoS attacks. Several researchers suggest that Machine Learning is a promising technique to fight Denial of Service attacks and they focus on using supervised learning methods to prove their theory. Using Classification to detect Denial of service attacks is expected to succeed especially since it can perform well during the training and testing phases but it wasn’t tested in real-life scenarios. In this paper we built seven different classifiers based on the CSE-CIC-IDS2018 dataset and tested them in the OMNET++ simulation environment since it is not possible to perform such tests on a real network. We found a large gap between the theoretical accuracy and the resulting one within the simulation which can be caused due to the Covariate shift problem. Tradition Classification might not be suitable to solve this problem. Other models were suggested to be tested in future studies.