nGraph-HE2: A High-Throughput Framework for Neural Network Inference on Encrypted Data
This work addresses the need for secure, high-throughput inference in domains like healthcare or finance by enabling encrypted data processing with minimal code changes, though it builds incrementally on prior frameworks.
The paper tackles the problem of enabling privacy-preserving neural network inference on encrypted data by extending nGraph-HE to support standard pre-trained models with native activation functions, achieving state-of-the-art throughput of 1,998 images/s on CryptoNets and evaluating MobileNetV2 on ImageNet with 60.4%/82.7% top-1/top-5 accuracy at 381 ms/image.
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.