SecretFlow is a unified framework for privacy-preserving data analysis and machine learning, developed by Ant Group’s SecretFlow team and published on GitHub with about 2,700 stars. It lets organizations train models and run analytics on data they cannot pool or reveal to each other.
SecretFlow layers cryptographic protocols, including secure multi-party computation, homomorphic encryption, differential privacy and trusted execution environments, behind a device abstraction: algorithms are written once as flows over plain and secret devices and then executed across horizontally or vertically partitioned data. Federated learning, split learning and private set intersection are built in, along with a workflow layer covering data processing, training and hyperparameter tuning. The framework is Apache-2.0 licensed and used in production privacy-computing platforms; development continues into 2026. The team maintains tutorials, a contribution map and good-first-issue tracks, and the framework underpins Ant Group’s production privacy-computing offerings used with financial and healthcare partners. Benchmarks and deployment guides for multi-party clusters are provided in the documentation, covering both simulation and real cross-silo setups.
SecretFlow is completely free and open source under the Apache-2.0 license. There are no paid tiers for the framework itself; you deploy it on your own infrastructure.
SecretFlow is for banks, healthcare organizations, governments and data teams that need joint machine learning and analytics without exposing raw data to partners.