Detecting Distributed Denial Of Service Attacks (DDoS) Using Machine Learning Models

Abstract

The digital landscape of today’s world is vulnerable to the widespread threat of Distributed Denial of Service (DDoS) attacks. These attacks have the potential to seriously damage businesses’ finances and reputations by interfering with the availability of internet services. Traditional methods of DDoS mitigation, such as rule-based approaches, struggle to keep up with the evolving nature of attacks. In this paper, I have trained and tested several supervised machine learning algorithms for the identification of DDoS attacks to determine the most effective one. I explore the depths of DDoS, obtaining and adjusting a dataset-utilizing principal component analysis (PCA) to reduce the number of features in the model from 80 to 20 while preserving 90% variance in our dataset. By reducing unnecessary features, PCA allowed us to have higher model accuracy and training speed. Overall, the Random Forest model trained with PCA had the best results, obtaining 99.9% accuracy, precision, and recall. The proposed approach exhibits encouraging results, demonstrating its potential to improve DDoS attack detection and thus reinforce network security.