Project created based on the paper Melissa Chase, Ran Gilad-Bachrach, Kim Laine, Kristin Lauter and Peter Rindal on Private Collaborative Neural Network Learning: link
Privacy is added by adding some randomness to the gradient and limiting the amount of impact a single data point in the dataset can do. The model was switched to a simpler model as some problems were encountered with the neural network framework.
Authors: Arnar Hardarson, Christian Koller Nielsen, Steffen Skovsgaard Jensen