This study investigates the use of ensemble machine learning methods in the detection of anomalies from blockchain transactions, including fraudulent transactions, in the Ethereum cryptocurrency network. A dataset of 9,840 Ethereum transactions labelled as fraudulent or legitimate was used to evaluate the performance of four machine learning models: Decision Tree, Random Forest, CatBoost, and XGBoost classifiers. Among these, the XGBoost classifier produced the best accuracy rate of 99.65%, in addition to better precision, recall, and F1-scores, exhibiting its excellent potential for effectively identifying fraudulent transactions. The findings reinforce XGBoost’s potential as a trusted mechanism for anomaly detection in blockchain transactions, which proves to be superior to the Decision Tree and Random Forest models. This research highlights the pivotal contribution of sophisticated machine learning algorithms in improving fraud detection in blockchain technology, offering a scalable and feasible solution to improve cybersecurity in the cryptocurrency space.