Privacy-Preserving Computation: Innovations and Practices at Jiànxìn Jīnke

Guest Speaker: Huo Yuguang, Senior Algorithm Researcher, Jiànxìn Jīnke

Editor: Song Ye, BoYu Technology

Platform: DataFunTalk

Introduction: The sharing focuses on industry applications, presenting Jiànxìn Jīnke's explorations and attempts in the field of privacy‑preserving computation.

Background: With data volume projected to increase from 7.6 ZB in 2018 to 48.6 ZB in 2025 in China, data leakage risks rise. Since GDPR’s enforcement in 2018, regulators have fined €272.5 million for violations. China has issued the Personal Information Protection Law and Data Security Law, emphasizing privacy protection.

Innovation Practice Path: Jiànxìn Jīnke has accumulated innovative practices in privacy‑preserving computation, developing a platform based on the FATE framework, adding proprietary functions, and integrating management and future data‑operation ecosystem capabilities.

The platform’s audit function combines blockchain for immutable evidence, ensuring traceability and reliability of multi‑party modeling processes.

Privacy‑Preserving Computation Platform:

1. Construction Principles and Ideas – The platform follows the enterprise architecture of China Construction Bank, integrating federated learning, MPC, and homomorphic encryption.

2. Application Paradigms – Three paradigms: joint query (protecting query results and conditions, i.e., Private Information Retrieval), joint computation (protecting data and computation methods), and joint modeling (building a shared machine‑learning model while preserving each party’s data privacy).

3. Platform Details – Built on FATE, enhanced with Jiànxìn Jīnke’s features, supporting management, data‑operation, and blockchain‑based evidence.

Application Pilot – Integrated Joint Marketing Model: A vertical federated tree model combines sales data from China Construction Bank with product features from Jiànxìn Fund, achieving a 34% improvement in targeting the top 5% customer segment compared with single‑party modeling.

Industry Leadership: Jiànxìn Jīnke participates in standards drafting, joins the FATE‑TSC, collaborates with UnionPay, WeBank, and others to create a financial‑industry branch of FATE, and contributes to privacy‑computing alliances and research projects.

Rapid Innovation Workshop: An open‑innovation mechanism aggregates scenarios, talent, data, and technology to build financial‑sector infrastructure, aiming to ensure security and stability of the national financial system.

Thank you for listening.