Privacy in AI

I was required to develop a privacy aware Artificial Intelligence Model for dealing with sensitive data such as medical records. This algorithm would need to protect user data from Membership Inference Attacks on the algorithms datasets.

A Deep Convolutional Neural Network classification algorithm was developed for the NIH Chest x-ray dataset using Python and TensorFlow.

Differential Privacy was used to apply privacy to the algorithm, which adds controlled noise to provide plausible deniability to the dataset.