High-Resolution Convolutional Neural Networks on Homomorphically Encrypted Data via Sharding Ciphertexts
This work enables secure, efficient deep learning inference on encrypted high-resolution data, addressing privacy concerns in domains like healthcare and finance, though it builds incrementally on existing encryption and neural network techniques.
The paper tackled the problem of privately evaluating deep convolutional neural networks on high-resolution encrypted data by extending methods to handle larger images and channels, achieving 80.2% top-1 accuracy on ImageNet and 98.3% on CIFAR-10 with speedups of 4.6-6.5x.
Recently, Deep Convolutional Neural Networks (DCNNs) including the ResNet-20 architecture have been privately evaluated on encrypted, low-resolution data with the Residue-Number-System Cheon-Kim-Kim-Song (RNS-CKKS) homomorphic encryption scheme. We extend methods for evaluating DCNNs on images with larger dimensions and many channels, beyond what can be stored in single ciphertexts. Additionally, we simplify and improve the efficiency of the recently introduced multiplexed image format, demonstrating that homomorphic evaluation can work with standard, row-major matrix packing and results in encrypted inference time speedups by $4.6-6.5\times$. We also show how existing DCNN models can be regularized during the training process to further improve efficiency and accuracy. These techniques are applied to homomorphically evaluate a DCNN with high accuracy on the high-resolution ImageNet dataset, achieving $80.2\%$ top-1 accuracy. We also achieve an accuracy of homomorphically evaluated CNNs on the CIFAR-10 dataset of $98.3\%$.