AI-Assisted Optimization of Hybrid RBAC-ABAC Access Control for Blockchain-Based Electronic Health Record Systems

Abstract

The proliferation of digital healthcare infrastructure has necessitated robust mechanisms for managing Electronic Health Records (EHRs) that balance security, privacy, and accessibility. Blockchain technology has emerged as a promising paradigm for EHR management, offering inherent properties of immutability, transparency, and decentralized trust distribution. However, the implementation of fine-grained access control mechanisms within blockchain-based EHR systems introduces substantial computational overhead and economic costs associated with blockchain transaction processing.

              Hybrid access control models that synergistically combine Role-Based Access Control (RBAC) and Attribute-Based Access Control (ABAC) have demonstrated enhanced flexibility and security compared to singular approaches. Nevertheless, these hybrid implementations predominantly rely on static policy evaluation mechanisms that exhibit significant inefficiencies when subjected to the dynamic, high-volume workloads characteristic of modern healthcare environments. The static nature of policy evaluation fails to capitalize on recurring access patterns and contextual similarities inherent in healthcare workflows, resulting in redundant computational operations and suboptimal resource utilization.

              This research introduces a novel AI-assisted optimization framework specifically designed for blockchain-enabled EHR systems, enhancing hybrid RBAC-ABAC access control through machine learning-based policy optimization. The proposed framework employs a Random Forest ensemble classifier to predict cost-efficient policy execution pathways by analysing contextual attributes, temporal patterns, and historical access behaviour. Critically, the optimization layer operates entirely off-chain, ensuring that no additional privacy vulnerabilities are introduced and that blockchain overhead remains minimal while computational intelligence is maximized.

              Comprehensive formal security analysis demonstrates that the proposed system preserves all cryptographic guarantees of the baseline implementation, maintaining collision resistance below 2^-128 while ensuring access control correctness through order-independent logical conjunction properties. The AI optimization layer is architecturally isolated from sensitive medical data, operating exclusively on access metadata and contextual features.

              Experimental evaluation conducted on an Ethereum-based testbed with simulated realistic hospital workloads reveals substantial performance improvements. The optimized framework achieves an additional 12-18% reduction in gas consumption and a 15-25% reduction in access decision latency compared to non-optimized hybrid RBAC-ABAC implementations. Scalability analysis demonstrates sub-linear growth in computational cost as system load increases, contrasting favourably with the near-linear growth exhibited by baseline approaches.

              The experimental results confirm that integrating AI-driven optimization significantly enhances the scalability, economic feasibility, and practical applicability of blockchain-based healthcare data management systems. This work provides a validated foundation for next-generation intelligent healthcare information systems that leverage machine learning to optimize blockchain operations without compromising the security and privacy guarantees essential for medical data management.