Privacy and machine learning

This project aims at reproducing a model inversion attack against a deep classifier of hand written digits (mnist).

Reference paper: https://arxiv.org/abs/1911.07135

Before starting make sure you have make installed (if not you can launch directly the docker commands)

Launch container with the following commands in terminal:

1. Build the Docker image

make build

2. Run the Docker image

make run

3. Return logs to get the address of the jupyter lab page (with the token needed)

make logs

4. When finished you can close and remove the container

make stop_and_rm