A New Framework for Handling Incomplete Data and Enabling Secure Access Control in eHealth

This research bridges these two disconnected areas to provide a novel link between incomplete data and security in a medical context.</p><p dir="ltr">To address these challenges, this research proposes a knowledge-driven framework that integrates data reconstruction with adaptive access control. The first contribution introduces two complementary approaches for reconstructing incomplete healthcare datasets. The first extends a neighbourhood-based k-nearest neighbour imputation algorithm to improve the estimation of missing numerical and categorical values, while the second employs semantic similarity techniques and domain-specific knowledge repositories to infer missing textual information.

Together, these approaches improve the completeness, consistency, and semantic quality of reconstructed datasets, providing a more reliable basis for subsequent identification and authorisation decisions. The reconstructed datasets are evaluated through diabetes diagnosis prediction as a proxy for assessing whether the reconstructed attributes are sufficiently reliable to support downstream security-sensitive decision making.

Experimental results demonstrate improved predictive performance, indicating that the proposed methods produce datasets that increase confidence in the use of incomplete healthcare information for decision-making. The research further investigates the characteristics and distribution of missing data, distinguishing between genuinely missing and non-applicable values, and examines the influence of data sparsity on reconstruction performance and algorithm selection.</p><p dir="ltr">Building upon the reconstructed datasets, the second contribution develops a modular, generalised hybrid attribute-based access control model designed to address several limitations of existing approaches.

The model incorporates features including donor inheritance, zonal inheritance, time-based access control, dynamic policy adaptation, and the application of Zero Trust principles throughout the decision-making process. A real-time simulated healthcare environment was designed and implemented to model realistic clinical workflows and evaluate access control decisions under varying levels of data incompleteness. The simulation provides a practical and repeatable experimental platform that overcomes the ethical, privacy, and operational constraints associated with experimentation in live clinical systems.

The hybrid model has been deliberately defined to be as flexible as possible to allow it to accurately represent real world scenarios which are essential for maintaining realism and the ability when transferring to real systems. Healthcare datasets rarely contain complete ground truth describing how access control policies affected attribute visibility. Since real clinical datasets rarely expose both complete patient information and the corresponding access-control decisions, a simulated environment is required to systematically evaluate how varying permission policies interact with incomplete data while maintaining known ground truth.</p><p dir="ltr">The proposed framework demonstrates that combining reliable data reconstruction with adaptive hybrid access control improves the quality and trustworthiness of identification and authorisation decisions in healthcare settings characterised by incomplete information.

The modular architecture enables individual components to be extended or replaced, making the framework applicable beyond healthcare and providing a foundation for future research into generic hybrid access control models capable of operating across diverse domains. The findings show that lightweight, customisable hybrid access control can address several open challenges in existing models while supporting secure and adaptive decision-making in dynamic environments.</p>