Towards Deep Encrypted Training: Low-Latency, Memory-Efficient, and High-Throughput Inference for Privacy-Preserving Neural Networks

TL;DR

This paper develops optimized algorithms and a pipeline architecture for batched homomorphic encryption neural network inference, significantly improving speed and memory efficiency for privacy-preserving applications.

Contribution

It introduces novel algorithms and a pipeline design tailored for batched HE inference, enabling high-throughput and resource-efficient encrypted neural network processing.

Findings

Achieved 1.78x faster inference time for ResNet-20 on encrypted CIFAR-10 images.

Reduced memory usage by 3.74x compared to previous methods.

Demonstrated scalable encrypted inference for deeper ResNet-34 models.

Abstract

Privacy-preserving machine learning (PPML) has become increasingly important in applications where sensitive data must remain confidential. Homomorphic Encryption (HE) enables computation directly on encrypted data, allowing neural network inference without revealing raw inputs. While prior works have largely focused on inference over a single encrypted image, batch processing of encrypted inputs lags behind, despite being critical for high-throughput inference scenarios and training-oriented workloads. In this work, we address this gap by developing optimized algorithms for batched HE-friendly neural networks. We also introduced a pipeline architecture designed to maximize resource efficiency for different batch size execution. We implemented these algorithms and evaluated our work using HE-friendly ResNet-20 and ResNet-34 models on encrypted CIFAR-10 and CIFAR-100 datasets, respectively. For ResNet-20, our approach achieves an amortized inference time of 8.86 seconds per image when processing a batch of 512 encrypted images, with a peak memory usage of 98.96 GB. These results represent a 1.78x runtime improvement and a 3.74x reduction in memory usage compared to the state-of-the-art design. For the deeper ResNet-34 model, we achieve an amortized inference time of 28.14 on a batch of 256 encrypted images using 246.78GB of RAM