The Tencent Angel PowerFL is a fast-growing platform for privacy-preserving computation, which provides a full-stack solution for achieving the great potential in federated learning and analytics. The Angel PowerFL platform enables multiple participants to jointly build machine learning models while keeping their data confidential from each other, and it has superior performance in terms of security, efficiency and reliability.
In this talk, we will introduce a novel federated communication framework, which is built on Apache Pulsar. We use Apache Pulsar as the fundamental component to build the secure communication channels, upon which a federation SDK is built as the communication protocol for multiparty privacy-preserving computation. We will also show that our framework can support real-time federated model serving and applications.