AWS Clean Rooms New2023年〜
A privacy-preserving data collaboration service that enables multiple organizations to jointly analyze data without sharing it
What It Does
AWS Clean Rooms is a privacy-preserving service that allows multiple companies or organizations to jointly analyze data without directly sharing their raw data. Each participant keeps their data in their own AWS account and can only run pre-agreed analysis rules (queries). Results are returned as aggregate values only, with no access to individual records.
Use Cases
Clean Rooms is used for ad effectiveness measurement between advertisers and media companies (conversion analysis without sharing personal data), joint clinical data analysis between pharmaceutical companies, collaborative marketing analysis of purchase data between retailers and manufacturers, and data matching for fraud detection between financial institutions.
Everyday Analogy
Think of it like a sealed ballot box. Each company puts their data (ballots) into a sealed box (Clean Rooms). Counting (analysis) happens inside the box, and only the results (aggregate totals) come out. Nobody can see who put in which ballot (individual data). You can understand overall trends while protecting data privacy.
What Is Clean Rooms?
AWS Clean Rooms is a service for multiple organizations to safely perform joint data analysis. In business, there are situations where you want to combine another company's data with your own for analysis. However, sharing raw data directly is difficult due to privacy and data security concerns. Clean Rooms provides a mechanism to achieve joint analysis without sharing data.
How Collaboration Works
In Clean Rooms, participants first create a "collaboration." Each participant registers their data stored in S3 as a "configured table" and controls which columns can be used for analysis and what types of aggregations (SUM, COUNT, AVG, etc.) are allowed. The party running the analysis can only execute permitted queries and cannot extract individual records.
Privacy Protection Mechanisms
Clean Rooms includes multiple privacy protection mechanisms. Analysis rules ensure only permitted aggregate queries can be executed. Output constraints let you set minimum row counts for results, preventing the risk of identifying individuals from small record sets. Additionally, Clean Rooms ML can apply differential privacy, adding statistical noise to make individual identification even more difficult. To understand the technical background of privacy protection mechanisms, reference books (Amazon) are helpful.
Getting Started
Create a collaboration in the Clean Rooms console and invite participants. Each participant registers their S3 data as configured tables and sets analysis rules (permitted query types, aggregation functions, output constraints). When the analyzing participant runs a query, aggregate results are returned according to the rules. We recommend starting with a test collaboration between two departments within your organization.
Things to Watch Out For
- Clean Rooms pricing is usage-based on the amount of data processed by queries. There are no additional charges for storing the data itself.
- Improperly configured analysis rules can lead to unintended data leakage or, conversely, prevent necessary analysis. Design rules carefully.
- All participants need AWS accounts. Collaborating with partners who don't use AWS requires separate arrangements.
