“‘latex This paper introduces ”SmartEdge,” a novel approach for optimizing the processing and analysis of Internet of Medical Things (IoMT) data. Through systematic preprocessing, advanced feature selection, controlled sampling, and innovative model aggregation, SmartEdge enhances performance and efficiency while minimizing computational overhead. The continuous improvement loop ensures adaptability to new data and operational conditions, promising significant advancements in IoMT ecosystem management. Our feature engineering on the CICIoMT2024 dataset emphasizes dimensionality reduction to improve computational efficiency in edge IoMT security and cyber attack detection. Techniques like Principal Component Analysis (PCA), feature selection, and embedding methods reduce the dataset’s dimensionality by 95%, drastically decreasing computational load and enabling real-time processing on resource-limited edge devices. This streamlining facilitates faster and more efficient machine learning model deployment, significantly boosting cyber attack detection and prevention in IoMT environments. Focusing on cyber resilience, we trained 10 edge models in just 55 seconds, with near-zero aggregation time due to an efficient method requiring no additional training or predictions. Using a small data set, five top features, and PCA with two components, we achieved high-performance results, with each model and the global model reaching accuracy above 0.99.