Keywords: Fully Homomorphic Encryption, Multi-Objective Co-Evolutionary Search, RNS-CKKS
TL;DR: Automated adaption of CNNs to the RNS-CKKS FHE scheme by jointly evolving polynomial activations (EvoReLUs) and searching for placement of bootstrapping operations.
Abstract: Secure inference of deep convolutional neural networks (CNNs) was recently demonstrated under the RNS-CKKS fully homomorphic encryption (FHE) scheme. The state-of-the-art solution uses a high-order composite polynomial to approximate non-arithmetic ReLUs and refreshes zero-level ciphertext through bootstrapping. However, this solution suffers from prohibitively high latency, both due to the number of levels consumed by the polynomials ($47\%$) and the inference time consumed by bootstrapping operations ($70\%$). Furthermore, it requires a hand-crafted architecture for homomorphically evaluating CNNs by placing a bootstrapping operation after every Conv-BN layer. To accelerate CNNs on FHE and automatically design a homomorphic evaluation architecture, we propose AutoFHE: Automated adaption of CNNs for evaluation over FHE. AutoFHE exploits the varying sensitivity of approximate activations across different layers in a network and jointly evolves polynomial activations (EvoReLUs) and searches for placement of bootstrapping operations for evaluation under RNS-CKKS. The salient features of AutoFHE include: i) a multi-objective co-evolutionary (MOCoEv) search algorithm to maximize validation accuracy and minimize the number of bootstrapping operations, ii) a gradient-free search algorithm, R-CCDE, to optimize EvoReLU coefficients, and iii) polynomial-aware training (PAT) to fine-tune polynomial-only CNNs for one epoch to adapt trainable weights to EvoReLUs. We demonstrate the efficacy of AutoFHE through the evaluation of ResNets on CIFAR-10 and CIFAR-100 under RNS-CKKS. Experimental results on CIFAR-10 indicate that in comparison to the state-of-the-art solution, AutoFHE reduces inference time (50 images on 50 threads) by 1,000 seconds and amortized inference time (per image) by $28\%$ and $17\%$ for ResNet-20 and ResNet-32, respectively.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics
Submission Guidelines: Yes
Please Choose The Closest Area That Your Submission Falls Into: Social Aspects of Machine Learning (eg, AI safety, fairness, privacy, interpretability, human-AI interaction, ethics)