Sessions

Breaking the DRAM Size Wall for DNN Inference and Homomorphic Encryption

Abstract

The proliferation of machine learning services in the last few years has raised privacy concerns. Homomorphic encryption (HE) enables inference using encrypted data but it incurs 100x-10,000x memory and runtime overhead. Secure deep neural network (DNN) inference using HE is currently limited by computing and memory requirements, with frameworks requiring hundreds of gigabytes of DRAM to evaluate small models. To overcome these limitations, we explore the feasibility of leveraging hybrid memory systems comprised of DRAM and persistent memory subsystems. In particular, we explore the recently-released Intel® Optane™ PMem to run large DNNs such as MobileNetV2 (in its largest variant) and ResNet-50 for the first time ever. We present an in-depth analysis of the efficiency of the executions with different hardware and software configurations. Our results conclude that DNN inference using HE yields friendly access patterns for this memory configuration, yielding efficient executions.

Speakers

Dr. Antonio J. Peña

Dr. Antonio J. Peña holds a BS+MS degree in Computer Engineering (2006), and MS and PhD degrees in Advanced Computer Systems (2010, 2013), from Universitat Jaume I de Castelló, Spain. He is currently a Sr. Researcher at the Barcelona Supercomputing Center (BSC), where he leads the “Accelerators and Communications for HPC” team.