Alibaba Mama Secure Data Hub: Cloud Architecture and Privacy-Preserving Advertising

In the advertising ecosystem, the secure and compliant use of data is a critical challenge. Alibaba Mama’s Secure Data Hub (SDH) addresses this by providing a privacy‑enhanced data clean‑room that enables cross‑domain data collaboration without exposing raw user information.

The platform leverages multi‑party computation (MPC), federated learning (FL) and differential privacy (DP) to offer brand‑level, consistent data decision capabilities while protecting user privacy.

Technical Architecture

SDH’s core capabilities include metadata‑level access control, encrypted join operators based on ECDH, secret‑sharing based inequality operators, and a suite of encrypted logical, relational and arithmetic operators. The system runs on a Flink‑based execution engine that validates plan legality and rewrites operators for secure execution.

Network communication is protected by a hybrid asymmetric‑symmetric encryption scheme, and distributed hash‑join and batch‑compressed data transfer optimize performance for billions of records.

SDH also provides a unified query API that accepts MPC‑SQL or invokes built‑in marketing analysis components, simplifying integration for advertisers.

Deployment

SDH can be deployed on Alibaba Cloud, third‑party clouds or private clouds via a serverless Kubernetes cluster, supporting elastic scaling and pay‑as‑you‑go pricing.

Application Cases

1. Full‑domain consumer asset analysis for a dairy brand, achieving >30% improvement in asset penetration, conversion and ROI.

2. Cross‑domain advertising effectiveness tracking, enabling secure user identification, joint modeling and multi‑channel conversion measurement for over ten leading brands.

Conclusion

SDH demonstrates how privacy‑preserving technologies can unlock data value in large‑scale advertising, offering a “usable but invisible” data flow paradigm that balances compliance with analytical power.