Enabling Homomorphic Encryption of Deep Neural Network Models and Datasets in Production Environments

Introduction

Deep learning (DL) is widely used to solve classification problems previously unchallenged, such as face recognition, and presents clear use cases for privacy requirements. Homomorphic encryption (HE) enables operations upon encrypted data, at the expense of vast data size increase.

Current Limitations

RAM sizes currently limit the use of HE on DL to severely reduced use cases. Recently emerged persistent memory technology (PMEM) offers larger-than-ever RAM spaces, but its performance is far from that of customary DRAM technologies.

Project Goals

This project aims at sparking a new class of system architectures for encrypted DL workloads by eliminating or dramatically reducing data movements across memory/storage hierarchies and network, supported by PMEM technology, overcoming its current severe performance limitations.

Objectives

HomE intends to be a first-time enabler for:

1. The encrypted execution of large models that do not fit in DRAM footprints to execute local to accelerators.
2. Hundreds of DL models to run simultaneously.
3. Large datasets to be run at high resolution and accuracy.

Research Scope

Targeting these ground-breaking goals, HomE enters into an unexplored field resulting from the innovative convergence of several disciplines, where wide-ranging research is required in order to assess current and future feasibility.

Main Challenge

Its main challenge is to develop methodology capable of breaking through the existing software and hardware limitations.

Proposed Approach

HomE proposes a holistic approach yielding highly impactful outcomes that include:

• Novel comprehensive performance characterisation.
• Innovative optimisations upon current technology.
• Pioneering hardware proposals.

Potential Impact

HomE can spawn a paradigm shift that will revolutionise the convergence of the machine learning and cryptography disciplines, filling a gap of knowledge and opening new horizons such as DL training on HE, currently too demanding even for DRAM.

Conclusion

HomE, based on solid evidence, will unveil the great unknown of whether PMEM is a practical enabler for encrypted DL workloads.