<![CDATA[This study investigates the effectiveness of homomorphic encryption (HE) in preserving patient privacy within Electronic Health Records (EHR) systems while enabling essential operations such as search, aggregation, and analytics. The core research question addresses how HE impacts system performance and resource utilization compared to conventional encryption methods. Using a prototype built on OpenEMR and a synthetic dataset of 1,000 outpatient records, we conducted three experimental scenarios: (1) single-record encryption–decryption, (2) batch aggregation (SUM and AVG) at batch sizes of 100, 500, and 1,000 records, and (3) ciphertext filtering based on a clinical threshold (glucose ≥ 200 mg/dL) across 1,000 records. Metrics collected included latency (ms), CPU usage (%), memory consumption (MB), and throughput (records/s). Statistical analysis via paired t-tests confirmed significant performance differences (p < 0.01) between HE and plaintext modes in all scenarios. Results demonstrate that HE introduces substantial overhead: single-record operations incurred average latencies of 115–121 ms versus 0.8–1.2 ms for conventional encryption; batch aggregation required 200–2,000 ms for HE compared to 1–12 ms for plaintext; and ciphertext filtering averaged 153 ms versus 0.8 ms. CPU usage increased from ~5 % to ~35 %, and memory from ~50 MB to ~210 MB under HE, while throughput dropped from hundreds to thousands of records per second to fewer than 1–7 records/s. Despite these costs, HE maintained 100 % accuracy in all operations. We conclude that HE is most suitable for scheduled batch processing in EHR environments that tolerate higher latency, rather than real-time clinical applications. Future work should explore optimization techniques, alternative HE schemes, and hybrid models to balance privacy and performance.]]>
Copyrights © 2025
