In the Fourth Industrial Revolution era, cybersecurity confronts an ever-evolving threat landscape with millions of distinct risks. Conventional detection methods prove inadequate as new vulnerabilities emerge incessantly, compounded by the exponential data generation rate. Machine learning offers a solution by enhancing data analysis, behavior prediction, and risk identification capabilities. This study explores advanced anomaly detection in secure cloud networks. The research aims to augment cloud anomaly detection through innovative methods: diverse feature set creation, randomized training data generation, and multi-algorithm evaluation. Systematically designing, implementing, and evaluating an improved detection system, using the UNSW-NB15 dataset to create six diverse training sets simulating varied attack scenarios. Comprehensive feature sets derived from various network aspects are defined. Six anomaly detection algorithms are rigorously tested, and evaluated based on accuracy, precision, recall, and F1-score, with random forest consistently outperforming others, achieving approximately 97.82% across Feature Set 1 and Feature Set 2. This research underscores machine learning’s crucial role in fortifying cloud network security against evolving cyber threats.