DeepSecure: Scalable Provably-Secure Deep Learning

TL;DR

DeepSecure introduces a scalable, privacy-preserving framework for deep learning using optimized garbled circuits, significantly improving throughput and efficiency while maintaining security in distributed data scenarios.

Contribution

It is the first to combine scalable secure deep learning with optimized garbled circuits and pre-processing techniques for enhanced performance.

Findings

Achieves over 58-fold throughput improvement compared to prior solutions.

Provides up to two orders-of-magnitude runtime reduction with pre-processing.

Enables secure delegation of computations to third parties in embedded environments.

Abstract

This paper proposes DeepSecure, a novel framework that enables scalable execution of the state-of-the-art Deep Learning (DL) models in a privacy-preserving setting. DeepSecure targets scenarios in which neither of the involved parties including the cloud servers that hold the DL model parameters or the delegating clients who own the data is willing to reveal their information. Our framework is the first to empower accurate and scalable DL analysis of data generated by distributed clients without sacrificing the security to maintain efficiency. The secure DL computation in DeepSecure is performed using Yao's Garbled Circuit (GC) protocol. We devise GC-optimized realization of various components used in DL. Our optimized implementation achieves more than 58-fold higher throughput per sample compared with the best-known prior solution. In addition to our optimized GC realization, we introduce a set of novel low-overhead pre-processing techniques which further reduce the GC overall runtime in the context of deep learning. Extensive evaluations of various DL applications demonstrate up to two orders-of-magnitude additional runtime improvement achieved as a result of our pre-processing methodology. This paper also provides mechanisms to securely delegate GC computations to a third party in constrained embedded settings.