Medical organizations require cloud-based machine learning frameworks that provide data protection and regulatory compliance to work with healthcare-related sensitive information and perform sophisticated analytics. The research delivers a fortified system that uses the MIMIC-III dataset and implements AES-256 encryption, TLS encryption, and AWS security components IAM and VPC to provide complete data security and HIPAA regulatory compliance. Advanced privacy-preserving methods such as differential privacy, homomorphic encryption, and federated learning are built into the system, which makes the data confidential and helps organizations meet regulatory standards. Using Keras for training a neural network model yields an outstanding AUROC of 0.82, surpassing all previous research findings. Future investigations concentrate on the expansion of dataset size, additional machine learning model exploration, and live anomaly detection system implementation to boost system performance for security and prediction exactness.