Concrete Numpy is an open-source set of tools which aims to simplify the use of fully homomorphic encryption (FHE) for data scientists.
With Concrete Numpy, data scientists can implement machine learning models using a subset of numpy that compiles to FHE. They will be able to train models with popular machine learning libraries and then convert the prediction functions of these models to FHE with concrete-numpy.
Links
For end users
Installation
The preferred way to use concrete-numpy is through docker. You can get the concrete-numpy docker image by pulling the latest docker image:
docker pull zamafhe/concrete-numpy:latest
To install Concrete Numpy from PyPi, run the following:
pip install concrete-numpy
You can find more detailed installation instructions in installing.md
A simple example: numpy addition in FHE
import concrete.numpy as hnp
def add(x, y):
return x + y
inputset = [(2, 3), (0, 0), (1, 6), (7, 7), (7, 1), (3, 2), (6, 1), (1, 7), (4, 5), (5, 4)]
compiler = hnp.NPFHECompiler(add, {"x": "encrypted", "y": "encrypted"})
print(f"Compiling...")
circuit = compiler.compile_on_inputset(inputset)
examples = [(3, 4), (1, 2), (7, 7), (0, 0)]
for example in examples:
result = circuit.encrypt_run_decrypt(*example)
print(f"Evaluation of {' + '.join(map(str, example))} homomorphically = {result}")For developers
Project setup
Installation steps are described in project_setup.md. Information about how to use Docker for development are available in docker.md.
Documenting
Some information about how to build the documentation of concrete-numpy are available in documenting.md. Notably, our documentation is pushed to https://docs.zama.ai/concrete-numpy/.
Developing
Some information about our terminology and the infrastructure of concrete-numpy are available in terminology_and_structure.md. An in-depth look at what is done in concrete-numpy is available in compilation.md.
Contributing
Information about how to contribute are available in contributing.md.
License
This software is distributed under the BSD-3-Clause-Clear license. If you have any questions, please contact us at [email protected].