This research dives into the world of cloud computing, focusing on the growing threat of Distributed Denial of Service (DDoS) attacks. These attacks can compromise the security of internet-connected systems, making them a significant concern. To tackle this, the study uses machine learning (ML) techniques to improve Intrusion Detection Systems (IDS), with a special focus on ensemble learning methodologies. It uses a variety of datasets, including a new one called DDoS-FD-22, which are split into training, testing, and validation subsets. The study introduces an innovative approach that integrates robust optimization algorithms (ROA) and ML techniques. It proposed methods like XGB-GA, RF-GA, and SVM-GA, which use robust optimization with a Tree-based Optimization Tool (TOT)-Genetic Programming. These models are trained on DDoS fusion datasets and optimized using ROAs, achieving high accuracy scores. Feature selection techniques are applied to identify class-dependent features and prevent overfitting. Impressively, our proposed model consistently achieves remarkable accuracy rates of 100%. Comparative analysis with other contemporary research papers demonstrates the superior performance of our proposed ML-ROA-based IDS system. Notably, the AdaBoost-GA ensemble model emerges as the most effective among tested models. In closing, the paper briefly explores the remaining challenges and future prospects for this research domain. It hints at the potential for advanced machine learning techniques and broader dataset testing, reinforcing its significance as an invaluable resource for newcomers in the realm of cloud security research.