A library for homomorphic statistics using tenseal
pip install git+https://github.com/kozzion/[email protected]
To update:
pip install git+https://github.com/kozzion/[email protected] --upgrade
To update with latest repo code:
pip install git+https://github.com/kozzion/[email protected] --upgrade --force-reinstall
The basic idea of this library is to a group of data holders to compute common statistics over their data without sharing any infomation on individual cases. There are three types of party involved: data holders, an agregator (which can also be a data holder) and a key holder. Due to the current limitations in HE scemes (lack of inversion, comparisons and square roots) some other information besides the test statistic and the degrees of freedom might be revealed to the decryptor.
Computation proceeds in 4 steps
- The key holder creates a public/ private key pair and publishes the public key
- The data holders encrypt their data using the public key and send it to the agregator
- The agregator uses the public key and the encrypted data to compute encrypted statistics and sends those to the key holder
- The key holder uses the private key to decrypt the statistics and publishes these to the data owners