Singapore has on Monday launched a new guide on federated learning to help businesses adopt privacy-enhancing technologies (PETs) for artificial intelligence (AI) development, as organizations seek ways to unlock the value of data while strengthening protection of personal information.
The Infocomm Media Development Authority (IMDA) said in a statement that Personal Data Protection Commission (PDPC) Singapore has unveiled a Guide on Federated Learning to help organizations understand and adopt Federated Learning.
It is noted that AI training and development requires centralizing data, but valuable data for AI development is often siloed, sensitive, or regulated, making it difficult to access and share.
Federated Learning offers a way to develop AI models by extracting the relevant data and insights.
Through the guide, businesses can expect to understand what Federated Learning is and how it can be used and adopted through case studies.
The guide also includes an adoption roadmap which helps businesses assess whether Federated Learning is suitable for their business needs, understand what is needed to adopt Federated Learning and how to design and configure a suitable Federated Learning solution. Recommendations for risk management are also provided within the guide.
PDPC has also updated the Guide on Synthetic Data Generation (SDG) with the latest industry research to advise organizations on the creation of artificial datasets.
Unlike Federated Learning which enables collaboration whilst keeping data stored locally, SDG creates artificial datasets that mirror the statistical properties of real data without containing personal information of individuals.
The updated guide expands on new generation methods and best practices to produce synthetic data and prevent re-identification. It also includes new case studies on organizations that have successfully generated synthetic data for practical applications.
The Federated Learning and SDG Guides help businesses assess which PET is suitable, adopt stronger safeguards, and unlock data for innovation, collaboration and AI development in a more privacy-preserving way.
The guides complement IMDA’s PET Sandbox, which provides a safe environment for organizations to explore PETs.
It is noted that PETs are tools that help organizations derive insights from personal data in a secure way.
They do this without exposing personal information, while ensuring data security.
PETs typically do this through two key approaches: enabling collaboration without sharing original data, using technologies such as Federated Learning and Trusted Execution Environments; and, enabling safer data sharing by protecting or transforming the data so that insights can be accessed without exposing the original dataset, leveraging technologies such as Synthetic Data Generation.
It is noted that IMDA’s PET Sandbox has seen wide industry participation and has facilitated numerous real-world use cases.
To date, 11 organizations from sectors such as finance, healthcare, construction, transport, and advertising tech have successfully implemented PETs within the Sandbox, with more expected to join in the pipeline.
To further encourage the adoption of PETs, IMDA is enhancing the PET Sandbox with new practical resources, which will allow organizations to explore and experience PETs through demonstrations of various PET tools before they test and implement their own use case within the Sandbox.