Designing a Privacy-Aware Framework for Ethical Disclosure of Sensitive Data

The increasing adoption of smart home healthcare ecosystems (SHHE) demands advanced privacy-preserving mechanisms to balance data utility with secure, ethical data disclosure. This thesis proposes a Privacy-Aware Authorization Framework that integrates a Dynamic Privacy Scoring Model (DPSM) and a Multi-Dimensional Dynamic Consent (MDDC) model within a decentralised smart contract infrastructure. This integration delivers context-aware, rule-enforced privacy decisions and decentralised, real-time consent enforcement with demonstrable accuracy, speed, and adaptability. The first phase of implementation achieved high consent enforcement accuracy (99.8-99.9%), response times within benchmark (2.45s at peak load), and successful support for 15,000 concurrent requests with 99.3% delivery. User evaluations confirmed strong usability (SUS score of 85.2) and high confidence in system transparency and control. These outcomes validate the robustness of the DPSM and MDDC in enabling compliant, efficient, and user-centric privacy governance. To further enhance adaptability and precision, a machine learning-driven Privacy Violation Prediction Model (PVPM) was introduced. This model supported system optimisation through proactive anomaly detection and data-driven risk insights. Its integration into the framework enabled dynamic tuning of access rules and consent policies, resulting in an F1-score of 0.98 and an AUC of 0.9976, confirming its value in mitigating evolving privacy threats and reducing manual intervention. This work contributes a scalable, adaptive privacy framework that harmonises mathematical scoring, user-centric consent, and intelligent automation. The proposed system establishes a benchmark for privacy-preserving design in SHHE while offering broader applicability to sectors requiring sensitive data control. Future research will explore advanced privacy-preserving techniques, including bio-authenticated dynamic consent, privacy-preserving federated learning, and quantum-resistant security models to address emerging threats while extending applications to multi-domain environments requiring sensitive data control.