nGraph-HE2: A High-Throughput Framework for Neural Network Inference on Encrypted Data

TL;DR

nGraph-HE2 advances privacy-preserving neural network inference by optimizing homomorphic encryption techniques, enabling real-world applications like MobileNetV2 on ImageNet with high throughput and accuracy.

Contribution

The paper introduces nGraph-HE2, extending previous work to support standard models with real activation functions and optimizing CKKS-based homomorphic encryption for better performance.

Findings

Achieved 1,998 images/sec throughput on CryptoNets.

Enabled homomorphic inference of MobileNetV2 on ImageNet.

Realized significant runtime speedups through multiple cryptographic and graph optimizations.

Abstract

In previous work, Boemer et al. introduced nGraph-HE, an extension to the Intel nGraph deep learning (DL) compiler, that enables data scientists to deploy models with popular frameworks such as TensorFlow and PyTorch with minimal code changes. However, the class of supported models was limited to relatively shallow networks with polynomial activations. Here, we introduce nGraph-HE2, which extends nGraph-HE to enable privacy-preserving inference on standard, pre-trained models using their native activation functions and number fields (typically real numbers). The proposed framework leverages the CKKS scheme, whose support for real numbers is friendly to data science, and a client-aided model using a two-party approach to compute activation functions. We first present CKKS-specific optimizations, enabling a 3x-88x runtime speedup for scalar encoding, and doubling the throughput through a novel use of CKKS plaintext packing into complex numbers. Second, we optimize ciphertext-plaintext addition and multiplication, yielding 2.6x-4.2x runtime speedup. Third, we exploit two graph-level optimizations: lazy rescaling and depth-aware encoding, which allow us to significantly improve performance. Together, these optimizations enable state-of-the-art throughput of 1,998 images/s on the CryptoNets network. Using the client-aided model, we also present homomorphic evaluation of (to our knowledge) the largest network to date, namely, pre-trained MobileNetV2 models on the ImageNet dataset, with 60.4\percent/82.7\percent\ top-1/top-5 accuracy and an amortized runtime of 381 ms/image.