Securing cloud infrastructures against malware has become increasingly critical due to the widespread adoption of cloud technologies. Traditional singlemodel detection approaches often fall short when dealing with the highdimensional, heterogeneous, and imbalanced nature of cloud-generated data. In this work, we propose AutoStack-MD, an automated stacked ensemble learning framework tailored for malware detection using tabular metrics data from cloud environments. Leveraging the capabilities of the AutoGluon AutoML framework, our system integrates advanced data preprocessing, feature selection, and hyperparameter optimization to construct a robust ensemble of diverse base models. To mitigate class imbalance, we apply the SMOTE technique and incorporate both bagging and stacking techniques to improve model generalization and predictive accuracy. Experimental evaluations on large-scale cloud datasets demonstrate that AutoStack-MD consistently outperforms baseline models in terms of accuracy and ROC AUC, confirming its effectiveness in real-world malware detection scenarios. This framework offers a scalable, adaptive, and automated solution that facilitates the deployment of intelligent security analytics within dynamic cloud infrastructures.