Deep Learning on Private Data

Emerging complex deep neural networks require vast amounts of data to achieve high precision. However, the information is often collected from user logs and personal data. In this article, we summarize recent cryptographic methodologies for provably privacy-preserving deep learning and inference.