A financial institution uses machine learning to predict loan defaults and wants to deploy the model in a secure and scalable manner. They are considering Amazon SageMaker for deployment. Which setup best adheres to security best practices while ensuring scalability?