Sphynx: A Deep Neural Network Design for Private Inference

Minsu Cho, Zahra Ghodsi, Brandon Reagen, Siddharth Garg, Chinmay Hegde

Abstract: Private inference (PI) methods involving deep networks enable cryptographically secure machine learning but incur severe latency costs. We outline how neural architecture searches are used to design networks with fewer nonlinearities, enabling efficient PI.