A Machine Learning-Based Privacy-Preserving Medical Diagnostic System

Abstract

In recent years, public awareness of healthcare has markedly improved, leading to a heightened demand for medical and health services. This trend is evident in the increased flow of outpatient visits, the rise in health check-ups, and the growing patient population. The surge in medical demand has resulted in exponential growth of sensitive health data, creating a critical conflict between data utility and privacy protection. This study proposes a machine learning-based privacy-preserving medical diagnostic system integrating end-to-end encryption, data desensitization, and adaptive learning algorithms. The cryptographic architecture employs SHA-256 for immutable data integrity verification, while synergistically combining AES-256 symmetric encryption (for efficient bulk data processing) with RSA-2048 asymmetric encryption (for secure key distribution), achieving both computational efficiency and quantum-resistant security. Experimental results demonstrate 98% diagnostic accuracy on multi-source clinical datasets, with hybrid encryption reducing latency by 37% compared to pure asymmetric approaches. The framework resolves the data-privacy dichotomy in healthcare and shows significant potential for enabling secure telemedicine in resource-limited regions through edge-device compatibility.

Ähnliche Arbeiten