Managing anonymised and secured medical big data from multiple sources

Abstract

The adoption of Electronic Health Records (EHRs) provides significant benefits to healthcare providers, but they continue to face persistent problems with data sharing and protection of privacy and system security. The current systems cannot connect patient information while maintaining privacy because healthcare systems generate massive volumes of textual and imaging data daily. Healthcare providers fail to deliver coordinated care because they do not have secure integration systems, which block them from accessing complete patient records.

This research creates a blockchain framework that allows healthcare organisations to potentially exchange medical information through protected patient identity preservation securely. The proposed framework addresses essential questions regarding patient identification methods that ensure privacy and blockchain configuration for various data types, system integration and scalability for national implementation, and security protection. The questions address the core problem that arises from balancing the individual privacy protection with making the required data available.

The framework consists of three elements that link cryptographic identification systems to blockchain optimisation methods and system integration procedures. The Global Identifier (GID) system generates permanent identifiers through SHA-256 cryptographic hashing, which enables recovery of the identifiers. Optimal block sizes for different medical data types were analysed and tested using Hyperledger Fabric v2.5.11. These components were then integrated and validated through limited proof-of-concept testing with one million synthetic records generated from Australian census data.

This research presents a theoretical framework demonstrating blockchain's potential for secure, privacy-preserving medical data integration through three key contributions. First, the Global Identifier system achieved 100% uniqueness across one million records while enabling secure recovery—solving the long-standing challenge of linking anonymised records. Second, the optimised blockchain architecture is designed to handle the 2,600:1 data volume ratio between imaging and text through differentiated storage (5MB blocks for text, 1GB for image metadata), with initial tests showing sub-second response times for metadata operations on a 2-node deployment. Third, the framework reduced synthetic data generation bias by 98% (from 13% to 0.1%), providing culturally authentic test data without compromising privacy.
While the framework achieved 234 transactions per second in limited empirical testing with metadata operations on a 2-node configuration, theoretical projections suggest potential for higher throughput in full-scale deployments. The distinction between empirically validated components (GID uniqueness, metadata performance, synthetic data generation) and theoretical projections (national-scale performance, comprehensive security resilience, etc) is maintained throughout this thesis.

Date of Thesis submission	2026
Original language	English
Awarding Institution	Western Sydney University
Sponsors	Australian Government Research Training Program Scholarship
Supervisor	Kenan Matawie (Supervisor), Nabil Giweli (Supervisor) & Mohammed Al-Zobbi (Supervisor)