UCLA Data Clean Room
Privacy-preserving CTR prediction with synthetic data
Back to Projects
HackathonResearch & Privacy
UCLA Data Clean Room
Privacy-preserving CTR prediction with synthetic data
Overview
Team BitBuilders built a secure Data Clean Room on Azure Confidential VMs with Intel TDX and SGX — enabling private data collaboration, synthetic data generation via GANs, and click-through rate prediction without exposing raw datasets across parties.
The Problem
Organizations need to collaborate on sensitive advertising and customer data without exposing raw datasets — especially for CTR modeling across parties.
The Solution
A Data Clean Room on confidential compute with remote attestation, containerized workloads, GAN-based synthetic data, and privacy-preserving CTR prediction.
Architecture
1
Confidential Compute
Azure CVMs with TEE isolation and vTPM for data-in-use protection.
2
Synthetic Data
GAN-generated datasets that preserve utility while limiting exposure.
3
Attestation
Quote verification and key return for trustable execution environments.
4
Analytics
CTR prediction and fidelity/utility evaluation across clean-room parties.
Technology
Azure Confidential VMIntel TDX / SGXGANDockerRemote AttestationCTR Modeling
Features
- Privacy-preserving data collaboration
- Synthetic data generation and evaluation
- Remote attestation on confidential VMs
- CTR prediction with multi-party data
- Containerized DCR workloads
Timeline
1
Research
Problem framing and institutional requirements
2
Build
Core implementation under hackathon timeline
3
Evaluation
Metrics, fidelity checks, and presentation delivery
Platform Screens
Ready to build institutional intelligence?
Let's discuss how Dreamskrin can architect your next platform.
Book Strategy Call