Hybrid Cloud Security Framework for E-Health Data Protection using Machine Learning and Advanced Encryption Techniques

Abstract

This study presents a hybrid cloud security framework for E-Health data protection that integrates machine learning-based behavioural analysis with advanced encryption techniques. With the use of Agile methodology, the system was developed to iteratively refine the proposed encryption algorithms in order to ensure adaptive functionality of the technique. A primary dataset made up of 15,001 user activity logs was collected from a cloud-based healthcare platform (Anderson Hospital) capturing both legitimate, suspicious or malicious behaviours. Furthermore, the dataset was then pre-processed using missing value imputation, min-max normalization and Principal Component Analysis (PCA) so as to optimize model training process. A Multilayer Perceptron (MLP) neural network was trained for the prediction of user sessions into three categories such as Legitimate, Suspicious or Malicious. The model achieved strong predictive continuous threat score performance with R² = 0.9946, MAE = 0.0689, and MSE = 0.0188, demonstrating a high predictive accuracy. For data protection, AES-128-bit encryption was used for routine access, while a hybrid Advanced Encryption Standard- Rivest–Shamir–Adleman (AES-256+RSA) approach secured high-risk scenarios. Then, the Experimental results show that the hybrid system provides robust security with acceptable processing overhead, ensuring confidentiality, integrity, and secure access control to sensitive health records. It significantly enhances security against key exchange vulnerabilities and interception attacks.The framework demonstrates the feasibility of real-time cloud-based E-Health data protection and provides a practical solution for safeguarding sensitive healthcare information.