Federated Learning (FL), which integrates privacy protection features, is being applied to the development of algorithms and data learning for fields such as healthcare and finance where mutual sharing or exposure is difficult, leading to an expansion of the ecosystem.

Intel announced on the 13th that it has approved Open Federated Learning (OpenFL) by participating in the incubation project of the LF AI & Data Foundation Technical Advisory Council.

OpenFL is an open-source framework for distributed AI that is a federated learning (FL) framework incorporating privacy features known as confidential computing. Intel developed and hosted this framework to enable data scientists to gain insights from heterogeneous, highly confidential, or regulated datasets while addressing data privacy issues.

Dr. Ibrahim Haddad (Dr. “This project is an innovative approach that enables organizations to collaboratively train machine learning models across multiple systems and data centers without sharing raw data, and it aligns perfectly with our mission to accelerate the growth and adoption of open-source AI and data technologies,” said Ibrahim Haddad, Executive Director of the LF AI & Data Foundation. “I look forward to successfully completing this project by collaborating with the talented individuals leading it.”

Through distributed machine learning approaches, data scientists can collaborate on analyses that achieve mutual benefits without exposing sensitive data or machine learning algorithms to other organizations. In industries such as healthcare, financial services, retail, and manufacturing, it is crucial to use federated learning to securely connect multiple systems and datasets, thereby removing obstacles to data aggregation for analysis and extracting valuable insights from the data.

Intel introduced OpenFL to the LF AI & Data Foundation alongside the University of Pennsylvania Perelman School of Medicine (Penn Medicine), VMware, and Flower Labs. Representatives from each company will join the Foundation to form the OpenFL Technical Steering Committee, which will establish a vendor-neutral ecosystem for the project and provide direction for its development. As the project is in the incubation phase, it is working with the LF AI & Data Foundation to lay the groundwork for its operational methods.

OpenFL is a framework for federated learning designed to be flexible, scalable, and secure. It enables enterprises to participate in multilateral machine learning processes without moving highly confidential or regulated data off-premises. Instead, algorithms process the data where it is held and centrally integrate the anonymized results. Data from any individual organization is not exposed to other organizations.

OpenFL combines hardware and software to further enable AI that protects privacy by using Intel Software Guard Extensions (Inte SGX), a hardware-based Trusted Execution Environment (TEE) for data centers, and The Gramine Project, a set of tools and infrastructure components for running unmodified applications on Intel SGX-based Confidential Computing Platforms.

Currently, OpenFL and Intel SGX open source integration is supported, and additional security features will be provided in the future. Organizations participating in the project can add integration with other TEE hardware to the project.