Homomorphic Encryption in Machine Learning as a Service (MLaaS)

Example of homomorphic encryption in machine learning inference as a service.

Train Model

train_model.py trains a regularized linear regression model on the California Housing Dataset, which is public data not requiring encryption but is nonetheless used here to demonstrate proof of concept.

Start Server

ml_server.py is a Flask application for the REST API. It accepts encrypted data from the client and returns a prediction without ever decrypting the data.

Run Client

client.py sends encrypted data to the server application and receives a response with a prediction that it decrypts.