Privacy Computing Enables Secure Medical Data Sharing and Analysis

Guest and Organizer : Dr. Wang Shuang, founder and chairman of NuoWei Technology, shared insights at a DataFunTalk event organized by Dalian University of Technology.

Company Overview : NuoWei Technology, a pioneer in medical privacy‑preserving computation, has built a secure, controllable privacy‑computing platform using federated learning, trusted execution environments, multi‑party computation, cryptography, and blockchain, earning national high‑tech certifications and multiple awards.

Background of Medical Data Privacy : Medical datasets (EHR, genomics, mobile health, public databases) require de‑identification before research use, yet re‑identification attacks using auxiliary information can recover sensitive details, as demonstrated by classic Sweeney attacks and recent studies showing residual privacy risks even after HIPAA Safe Harbor processing.

Limitations of Simple De‑Identification : De‑identification alone cannot guarantee anonymity, especially for high‑dimensional data like genomes; regulatory definitions distinguish anonymized data from merely de‑identified data, highlighting the need for stronger protection mechanisms.

Sandbox and Privacy‑Computing Solutions : Sandbox execution mitigates some de‑identification issues but struggles with multi‑source collaborative analysis, prompting the development of privacy‑computing frameworks that virtually fuse data while preserving model accuracy comparable to plaintext aggregation.

Technical Approaches : Privacy computing is categorized into three main techniques: federated learning (local computation with gradient/model exchange), trusted hardware (e.g., Intel SGX enclaves providing isolated execution), and cryptographic methods (multi‑party computation, homomorphic encryption). Each has distinct trust models, implementation difficulty, flexibility, and scalability.

Team Achievements : The team contributed early work on secure federated learning for nationwide medical networks, trusted execution environment standards, homomorphic encryption standards, and multi‑party computation integration, receiving Intel’s Outstanding Contribution Award and numerous academic publications.

Product Platform : NuoWei offers a comprehensive privacy‑computing platform deployable as appliances, software packages, containers, or cloud services. It supports diverse algorithms (logistic regression, XGBoost, neural networks), integrates third‑party SDKs, and provides blockchain‑based audit trails.

Application Scenarios : The platform enables multi‑center clinical research, epidemic monitoring, drug discovery, privacy‑preserving queries, whole‑genome analysis, and impact studies across hospitals and regions, with real‑world deployments in cross‑province gene analysis, nationwide cancer CDR sharing, and international Kawasaki disease studies.

Case Studies : Notable projects include a cross‑province multi‑center gene analysis system, a cancer data sharing network covering 60+ top hospitals in 24 provinces, and a global privacy‑preserving analysis system for rare diseases and infectious disease early‑warning, all demonstrating high accuracy and compliance.

Cloud AI Service : To address the need for large‑scale compute, NuoWei provides privacy‑preserving cloud AI services where data remains encrypted throughout storage and multi‑center analysis.

Conclusion : Privacy computing bridges the gap between data utility and confidentiality, offering scalable, secure solutions for medical big data while maintaining model performance and complying with regulatory requirements.