All Tools
P
OtherFreeOpen Source
PYSYFT
Perform data science on data you cannot see — privacy-preserving ML
Apache-2.0
ABOUT
ML teams often cannot access the most valuable training data because it is sensitive — patient records, financial transactions, personal data — and cannot leave its owner's infrastructure. PySyft lets data scientists write queries against remote datasets without ever seeing the raw data, using a combination of federated execution, differential privacy, and SMPC. The data never moves; the computation comes to the data.
INSTALL
pip install syftINTEGRATION GUIDE
1. Train a federated model across hospital patient records without any patient data leaving each hospital
2. Run differentially private statistical analyses on sensitive financial data with formal privacy guarantees
3. Build a multi-party computation pipeline where two companies jointly compute insights on their combined data
4. Delegate model training to a remote data owner's infrastructure with end-to-end encrypted computation
5. Create a privacy-preserving data science workflow that enforces access policies at the tensor level
TAGS
privacyfederated-learningdifferential-privacyencrypted-computationdata-sciencesecuritypython