AWS Clean Rooms ML のアイコン

AWS Clean Rooms ML New2023年〜

A service for building ML models within Clean Rooms while preserving privacy

About 3 min read

What It Does

AWS Clean Rooms ML is a service that enables multiple companies to jointly build machine learning models without directly sharing their data. Each company's data is used for analysis in a privacy-protected state, and raw data is never exposed to other parties. It enables safe data collaboration for use cases such as ad effectiveness prediction and lookalike audience modeling.

Use Cases

Clean Rooms ML is used when advertisers and publishers want to improve ad targeting accuracy without directly sharing customer data. It is also used when retailers and manufacturers combine purchase data to build demand forecasting models, and when financial institutions jointly train fraud detection models.

Everyday Analogy

Imagine multiple companies cooking together. Each company doesn't want to reveal their secret recipe (data) to others. Clean Rooms ML is a system where each company puts their ingredients into a sealed box, and only a robot chef (ML model) uses the contents to complete the dish. Everyone receives the finished dish (analysis results), but nobody can see anyone else's recipe.

What Is Clean Rooms ML?

AWS Clean Rooms ML is a service that extends AWS Clean Rooms to build machine learning models while preserving privacy. Traditionally, combining data from multiple companies to create ML models required gathering data in one place. With Clean Rooms ML, you can train and run inference on ML models in a secure environment without moving or sharing each company's data.

Using Lookalike Models

A key feature of Clean Rooms ML is the Lookalike Model. For example, an advertiser can use their list of high-value customers to find users with similar characteristics in a publisher's user base. Neither the advertiser's customer data nor the publisher's user data is exposed to the other party, enabling high-precision targeting while maintaining privacy.

Privacy Protection Mechanisms

Clean Rooms ML uses differential privacy and encryption techniques to prevent individual records from being identified. Collaboration rules (which data can be used and to what extent) are configured in advance, and queries or analyses that violate these rules are automatically blocked. Data owners always maintain control over their own data. To gain a deeper understanding of privacy protection mechanisms, you can also refer to specialized books (Amazon).

Getting Started

To get started with Clean Rooms ML, first create a collaboration in AWS Clean Rooms and define data usage rules with participating members. Next, select the ML model type (such as Lookalike Model) in the Clean Rooms ML settings and specify the training data source. Once model training is complete, you can export the results and apply them to marketing initiatives.

Things to Watch Out For

  • Clean Rooms ML is an add-on feature of AWS Clean Rooms, so you need to set up a Clean Rooms collaboration first.
  • Lookalike Model accuracy depends on the quantity and quality of training data, so ensure you have sufficient data volume.
  • Available regions are limited, so check supported regions in advance.
共有するXB!

Related Services

Amazon SageMaker iconAmazon SageMakerAn integrated platform for building, training, and deploying machine learning modelsAmazon S3 iconAmazon S3A scalable object storage serviceAWS Clean Rooms iconAWS Clean RoomsA privacy-preserving data collaboration service that enables multiple organizations to jointly analyze data without sharing it

Related Articles

Privacy-Preserving ML with AWS Clean Rooms ML - Build Models Without Sharing DataLearn how to build lookalike models with Clean Rooms ML, apply differential privacy, and leverage the results for ad targeting.

More in This Category

Amazon Bedrock iconAmazon Bedrock PopularA fully managed generative AI service providing access to leading foundation models via APIAmazon Comprehend iconAmazon Comprehend SpecializedA service that uses natural language processing to extract entities, sentiment, and key phrases from textAmazon Forecast iconAmazon Forecast DeprecatedA fully managed service that uses machine learning to forecast time series dataAmazon HealthLake iconAmazon HealthLake SpecializedA managed service for storing, transforming, and analyzing FHIR-compliant healthcare dataAmazon Kendra iconAmazon Kendra DeprecatedAn intelligent enterprise search service powered by machine learningAmazon Lex iconAmazon Lex SpecializedA service for building conversational interfaces using voice and textAWS Panorama iconAWS Panorama SpecializedA service that adds computer vision to existing IP cameras for edge-based video analysisAmazon Personalize iconAmazon Personalize DeprecatedBuild real-time personalized recommendations without machine learning expertise

Similar Articles and Services

Privacy-Preserving ML with AWS Clean Rooms ML - Build Models Without Sharing DataLearn how to build lookalike models with Clean Rooms ML, apply differential privacy, and leverage the results for ad targeting.Privacy-Enhanced Data Analytics - Secure Data Collaboration with AWS Clean RoomsLearn about privacy-enhanced data analytics with AWS Clean Rooms. Explore multi-organization data collaboration, access control through analysis rules, and differential privacy.AWS Clean RoomsA service that enables multiple organizations to perform joint analysis without exposing each other's raw data, balancing privacy protection with analyticsPrivacy-Preserving Data Collaboration with AWS Clean RoomsRun joint analysis across multiple companies without sharing or copying data. Learn about aggregation rules for preventing individual identification and Cryptographic Computing for encrypted analysis.