NIST Releases Open-Source Privacy-Preserving Tool
The new federated learning testbed helps organizations train machine learning models while safeguarding sensitive data.
Updated on Oct. 30, 2026 in Cybersecurity
Live Poll
Do you trust current privacy technologies to protect sensitive data while allowing it to be shared?
The NIST National Cybersecurity Center of Excellence has launched an open-source Privacy-Preserving Federated Learning (PPFL) testbed paired with a genomic dataset. This tool is designed to help users evaluate how privacy settings influence machine learning model performance.
Why it matters
The project offers a practical environment for testing how Privacy-Enhancing Technologies allow organizations to leverage sensitive data without compromising individual privacy. It builds on previous security insights gained during a 2025 public red-teaming exercise.
The PPFL Testbed enables training machine learning models either centrally or across federated sites using optional differential privacy configurations. The accompanying dataset includes over 18,000 accessions from the USDA Soybean Germplasm Collection.
The players
NIST National Cybersecurity Center of Excellence
The NCCoE is a division of the National Institute of Standards and Technology that focuses on providing practical cybersecurity solutions to organizations.
USDA
The U.S. Department of Agriculture manages the germplasm collection used to provide the genomic data for the testbed.
The details
The testbed allows users to conduct comparative analyses on how different privacy configurations and settings impact the final performance of their models. This initiative is a core component of the broader NIST Privacy-Enhancing Technologies (PETs) Testbeds program.
Timeline
A public red-teaming exercise for the technology occurred in 2025.
NIST officially released the PPFL Testbed and the curated genomic dataset on October 29, 2026.
The Tech Race
This release marks a significant progression within the NIST Privacy-Enhancing Technologies (PETs) Testbeds program. It directly addresses the industry-wide transition toward secure, decentralized data processing methods that replace traditional, high-risk centralized data silos.
Developers and organizations can now utilize these tools to improve machine learning workflows while maintaining strict data privacy standards. This allows for more secure data processing in sectors like healthcare and agriculture where sensitive information is common.
The takeaway
Privacy-Enhancing Technologies are becoming essential tools for organizations that need to balance model performance with strict data security requirements. Developers should explore these open-source testbeds to see how differential privacy settings can be integrated into their own machine learning pipelines.
What happens next
The NCCoE plans to release additional open-source testbeds, datasets, and reference architectures in future phases of the project.
Further reading
For additional context on how the government is securing digital infrastructure, visit the Cybersecurity section.
More information
Review the full PETs Testbeds project information page to access the tools.
Source note: This article includes information reported by NIST.
Live Poll
Do you trust current privacy technologies to protect sensitive data while allowing it to be shared?
