IEEE P3652.1 Federated Machine Learning Working Group Meeting continues work on standard-setting for Federated Learning, with a focus on scenario and demand classification and security test.

Curtains fall on the 4th IEEE P3652.1 (Guide for Architectural Framework and Application of Federated Machine Learning) Standard Working Group Meeting in Beijing. Focused on classifying the scenarios and demand for Federated Learning (FL), the meeting developed a plan for the security test and rating of FL, and deepened discussion on standard-setting for FL. Joining the meeting are 22 leading enterprises and research institutions namely Peking University, IEEE, WeBank, Sinovation Ventures, JD, China Telecom, Tencent, MI, Alibaba, YITU, Clustar, 4Paradigm, Huawei, VMWare, LogiOcean, SensesGlobal, Swiss Re, Intel, CETC BigData, Ant Financial Services Group, ChinaAMC, Fudata Technology.

The implementation of AI technology in reality encounters two main bottlenecks in recent years. On the one hand, “small data” possessed by many different enterprises fail to pool together and become complementary. On the other hand, increasing attention on data privacy and security has become a global trend. As an encrypted and distributed paradigm for Machine Learning, Federated Learning provides more possibilities to countering the dilemma of AI implementation by allowing all parties to co-build a model without disclosing their original data. Application of this emerging AI technology is already seen in various sectors including finance, healthcare, city management.

Launched by WeBank and approved last December, IEEE P3652.1 Program is the first in the world dedicated to standard-setting for AI Collaborative Technology Framework, in a bid to provide technical standards for the application of FL and a basis for cooperation in jointly building Federated ecosystem among all walks of life.

Building on the previous 3 meetings, the 4th meeting made progress by adopting a more detailed perspective in classifying Federated Learning under scenarios such as To B (Business), To C (Customer), To G (Government), developing a demand template for Federated Learning and a detailed plan for security test of FL, adding great substance to the standard-setting of FL, which serves as an important boost to the Draft of Standards for Federated Learning. A schedule was also formulated at the meeting: Draft of Standards is expected to be introduced in February 2020, and Official Standards in the first half of 2020. Since the launch of this program, over 30 internet giants, government institutions, enterprises and universities have been contributing to standard-setting in finance, technology, healthcare, education, etc. Members of the working group (with the right to vote) include WeBank, Tencent, JD, Intel, Huawei, China Telecom, MI, Eduworks, Sinovation Ventures, Squirrel AI, Clustar, 4Paradigm, CETC BigData, LogiOcean, SensesGlobal.

With data security and privacy protection in the spotlight, Federated Learning is expected to become the basis for the next generation of AI collaborative network, thus build trust between organizations and users in terms of data, and promote the adoption of technology for a good cause. The standard-setting efforts will further contribute to a standardized system for the application of FL in different sectors. Based on unified technical standards, all sectors of the society will surely work together to build an ecosystem for tapping the full potential of FL and a new path for the development of AI.