Concrete-ML is an open-source set of tools which aims to simplify the use of fully homomorphic encryption (FHE) for data scientists. Particular care was given to the simplicity of our Python package in order to make it usable by any data scientist, even those without prior cryptography knowledge. Notably, our APIs are as close as possible to scikit-learn and torch APIs to simplify adoption by our users.

• Links
• For end users
  • Installation.
  • Supported models.
  • Simple ML examples with scikit-learn.
• For developers
  • Project setup.
  • Documenting.
  • Developing.
  • Contributing.
• License

• documentation
• community website
• demos

The preferred way to use Concrete-ML is through Docker. You can get our Docker image by pulling the latest Docker image:

docker pull zamafhe/concrete-ml:latest

To install Concrete-ML from PyPi, run the following:

pip install concrete-ml

You can find more detailed installation instructions in installing.md

Here is a list of ML algorithms currently supported in this library:

• LinearRegression (sklearn)
• LogisticRegression (sklearn)
• SVM - SVC and SVR (sklearn)
• DecisionTreeClassifier (sklearn)
• RandomForest (sklearn)
• NeuralNetworkClassifier (skorch)
• NeuralNetworkRegressor (skorch)
• XGBoostClassifier (xgboost)

Torch also has its own integration for custom models.

A simple example which is very close to scikit-learn is as follows, for a logistic regression :

from sklearn.datasets import make_classification
from sklearn.model_selection import train_test_split

from concrete.ml.sklearn import LogisticRegression

# Create a synthetic dataset
N_EXAMPLE_TOTAL = 100
N_TEST = 20
x, y = make_classification(n_samples=N_EXAMPLE_TOTAL, class_sep=2, n_features=4, random_state=42)
X_train, X_test, y_train, y_test = train_test_split(
    x, y, test_size=N_TEST / N_EXAMPLE_TOTAL, random_state=42
)

# Fix the quantization to 3 bits
model = LogisticRegression(n_bits=3)

# Fit the model
model.fit(X_train, y_train)

# We run prediction on non-encrypted data as a reference
y_pred_clear = model.predict(X_test, execute_in_fhe=False)

# We compile into an FHE model
model.compile(x)

# We then run the inference in FHE
y_pred_fhe = model.predict(X_test, execute_in_fhe=True)

print("In clear  :", y_pred_clear)
print("In FHE    :", y_pred_fhe)
print("Comparison:", (y_pred_fhe == y_pred_clear))

# Output:
#   In clear  : [0 1 0 1 0 1 0 1 1 1 0 1 1 0 1 0 0 1 1 1]
#   In FHE    : [0 1 0 1 0 1 0 1 1 1 0 1 1 0 1 0 0 1 1 1]
#   Comparison: [ True  True  True  True  True  True  True  True  True  True  True  True
#   True  True  True  True  True  True  True  True]

We explain this in more detail in the documentation, and show how we have tried to mimic scikit-learn and torch APIs, to ease the adoption of Concrete-ML in this page dedicated to scikit-learn and in this page dedicated to torch.

Installation steps are described in project_setup.md. Information about how to use Docker for development are available in docker.md.

Some information about how to build the documentation of Concrete-ML are available. Notably, our documentation is pushed to https://docs.zama.ai/concrete-ml/.

Some information about the infrastructure of Concrete-ML and some of the core elements we use are available here. Notably, an in-depth look at what is done in Concrete-ML is available in onnx_use_for_compilation.md.

Information about how to contribute is available in contributing.md.

This software is distributed under the BSD-3-Clause-Clear license. If you have any questions, please contact us at [email protected].