The increase in the sophistication of DDoS attacks poses severe risks to availability, security, and performance in cloud computing environments. This SLR work will present a critical review of research papers involving the adoption of supervised learning, ensemble methods, and hybrid models for detecting and predicting DDoS attacks. It evaluates the influence of the ensemble techniques on detection accuracy and false positive rates, and it assesses the role of supervised, unsupervised, and reinforcement learning for attack prevention. The review discusses the major challenges of scalability, data management, and adap-tive response mechanisms and emphasizes that hybrid ML-rule-based approaches are potential solutions for robust and scalable cloud security.