Alibaba moves its latest privacy-preserving computing technology to open source

Alibaba DAMO Academy, a global research initiative of the Alibaba Group, announced that it has provided the open source community with the source code for its latest federated learning platform, FederatedScope, a comprehensive platform with easy-to-use packages.

With the advent of machine learning in the digital age, the collection of training data to build and improve AI models is under scrutiny as the process can create privacy concerns. To solve this problem, federated learning has emerged, a computational method that preserves privacy.

By coordinating microtask training across different end devices, intermediate training results, rather than raw user data, are sent back to the cloud server to mitigate privacy concerns. This still allows you to perform data analysis and machine learning tasks on end devices.

“By sharing our federated learning technologies with the open source community, we hope to facilitate research and industrial deployment of privacy-preserving computing in industries such as healthcare, which typically involve sensitive user data and require strict privacy regulations.”Bolin Ding, a researcher at the Alibaba DAMO Academy, said.

In addition, with a newly implemented event-driven platform, FederatedScope offers flexible support and comprehensive tools, including a rich collection of reference datasets, well-known model architectures, advanced federated learning algorithms, easy-to-use autotuning features, and customizations. friendly interfaces. These tools allow researchers and developers to quickly create and customize federated learning applications for specific domains. such as computer vision, natural language processing, speech recognition, graph learning, and recommendations.

For privacy in particular, the platform also offers state-of-the-art technologies, including differential privacy and multi-party computing to meet various privacy requirements.

“We believe that privacy-preserving computers are an important and important trend. The ability to train AI models without compromising privacy is critical, which is why we’ve dedicated a lot of resources to spur research into federated learning. We hope that by sharing our source code and technology platform, we will be able to support the global developer community and spur new innovation in this new field. »

By 2025, 60% of large organizations are expected to use one or more privacy-enhancing computing technologies, according to Gartner.

Earlier this year, the Alibaba DAMO Academy also released its predictions of the major trends that will shape the tech industry in the coming years, and one of them was privacy-preserving computers. Forecasts for the next three years We should “witness revolutionary improvements in the performance and interpretability of privacy-preserving computing.”

To learn more about Alibaba Cloud’s federated learning platform and technologies, visit and the dedicated GitHub page.