Ontology-Driven Attribute-Based Access Control with Optimized Martino Encryption for Secure Healthcare Cloud Data

Abstract

Ontologies are extensively utilized in various domains to model trust-based data access management. However, enforcing time-limited access permissions remains difficult and may lead to unauthorized access after the collaboration period expires. Initially, the Kohonen Self-Organizing Neural Network (KSONN) analyzes healthcare requests and classifies them as normal or malicious. Normal requests are then scheduled using the Prioritized Sorted Task Scheduling Algorithm (PSTSA), with task priority determined from the request context rather than from the user role alone. The fuzzy system considers the role, urgency, clinical criticality, impact, resource availability, and deadline proximity. Ontology-derived facts are mapped to eXtensible Access Control Markup Language (XACML) subject, resource, action, and environment attributes before policy evaluation. For example, ontology reasoning identifies the role of the user and the relationship to a patient record, and these values are supplied as subject and resource attributes to the XACML policy decision point. Cluster labels are assigned using only the class labels of the training samples mapped to each cluster. Test-set labels are not used during cluster formation or cluster-label assignment and are only used for the final performance evaluation. If the data is classified as normal, the job scheduling process is performed based on the PSTSA. Finally, patient privacy is protected using the proposed cryptographic method (MESBO-based Martino encryption), and the encrypted healthcare data are securely stored in the cloud. The framework is estimated using an 80:20 train-test split, and the reported results represent the average performance across ten independent experimental runs. The representative KSONN test run achieved 97.93% accuracy and 99.39% precision, whereas the ten-run mean performance is used for comparative evaluation.