Machine Learning Pipeline: Feature Selection and Adaptive Training for DDoS Detection to Improve Cloud Security

DDoS attacks are a concern in most distributed and cloud environments, and they can be a threat to any multi-cloud system. This research offers an innovative method to detect DDoS using adaptive machine learning techniques. The proposed methodology deploys a combination of algorithms, such as LightGBM, CatBoost, and XGBoost, with an overall accuracy of 99.32%, 99% specificity, and 99% sensitivity for most attack classes. In addition, the methodology addressed the challenges of the minority classes, where CatBoost had a recall of 85% for previously marginalized attacks. The results indicate the effectiveness of the proposed system across different DDoS attack types and traffic patterns, making it viable and effective for the protection of cyber security structures that operate in a multi-cloud system.