An Evaluation Framework for Machine Learning Methods in Detection of DoS and DDoS Intrusion

A distributed denial-of-service (DDoS) and DoS attack are the most devastating and expensive attacks among various cyber and network attacks [1] [2] . Coupled with the fact that launching such attacks could be relatively easy, it makes it a big problem in the realm of Security and Cyber Space in general. However, with the advent of advanced Artificial Intelligence / Machine Learning(AI/ML) methods and tools, we explore different research techniques and methodologies to find a better detection accuracy result and prevent many different kinds of Attacks and Intrusions. During the research process, we will address Analytical and Computational challenges, Feature Selection issues, and Machine Learning Models while paying particular attention to Feature Engineering by using Mutual Information and Principal Component Analysis in the feature construction process. Moreover, K-Nearest Neighbors, Decision Trees, Random Forests, and XGBoost for Classification are used. In General, this study will target to analyze the ability of these methods to detect DoS and DDoS attacks while also examining the capacity of the ways to distinguish between different kinds of these attacks. Finally, the research investigates and proposes a framework for simultaneous evaluation of different Machine Learning methods in detecting DoS and DDoS.