XMODOS: An Explainable Model for Denial of Service Attack Detection

This paper proposes a Machine Learning-based approach for detecting Denial-of-Service (DoS) attacks using different datasets for training and testing. The study evaluates the performance of the proposed approach on a unique dataset; thereafter, compares it to existing approaches in a bid to demonstrate its superiority in terms of accuracy, false-positive rate, and computational efficiency. Furthermore, the impacts of different Machine Learning algorithms and hypertuning configurations on the performance of the proposed approach are investigated. Also, via Feature Importance Analysis, this study examines the influence of each feature (present in the dataset) on our proposed model. The contributions of our work herein demonstrate the potentials of the proposed approach in detecting DoS attacks; and our research highlights the importance of employing Machine Learning in this domain. Future research directions have been suggested based on insights acquired from the investigations of different algorithms and hyperparameters.