MetaKernel: Enabling Efficient Encrypted Neural Network Inference through Unified MVM and Convolution
Proceedings of the ACM on Programming Languages · 09.10.2025 · 9 (OOPSLA2) · 1261-1288
DOI: 10.1145/3763095 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of the ACM on Programming Languages, online erschienen am 09.10.2025. Verfasst von Peng Yuan und 8 weiteren Autor:innen. Die Arbeit wurde bisher 3-mal zitiert (Crossref) und nennt 46 Literaturangaben. Gefördert von: National Key R&D Program of China.
Abstract (Originalsprache)
Practical encrypted neural network inference under the CKKS fully homomorphic encryption (FHE) scheme relies heavily on accelerating two key kernel operations: Matrix-Vector Multiplication (MVM) and Convolution (Conv). However, existing solutions—such as expert-tuned libraries and domain-specific languages—are designed in an ad hoc manner, leading to significant inefficiencies caused by excessive rotations. We introduce MKR, a novel composition-based compiler approach that optimizes MVM and Conv kernel operations for DNN models under CKKS within a unified framework. MKR decomposes each kernel into composable units, called MetaKernels , to enhance SIMD parallelism within ciphertexts (via horizontal batching) and computational parallelism across them (via vertical batching). Our approach tackles previously unaddressed challenges, including reducing rotation overhead through a rotation-aware cost model for data packing, while also ensuring high slot utilization, uniform handling of inputs with arbitrary sizes, and compatibility with the output tensor layout. Implemented in a production-quality FHE compiler, MKR achieves inference time speedups of 10.08×−185.60× for individual MVM and Conv kernels and 1.75×−11.84× for end-to-end inference compared to a state-of-the-art FHE compiler. Moreover, MKR enables homomorphic execution of large DNN models, where prior methods fail, significantly advancing the practicality of FHE compilers.
Förderung
- National Key R&D Program of China
Zitieren
APA
Yuan, P., Liu, Y., Lai, J., Li, L., Sui, T., Xiao, L., Zhang, X., Zhu, Q., & Xue, J. (2025). MetaKernel: Enabling Efficient Encrypted Neural Network Inference through Unified MVM and Convolution. Proceedings of the ACM on Programming Languages, 9(OOPSLA2), 1261-1288. https://doi.org/10.1145/3763095
Vancouver
Yuan P, Liu Y, Lai J, Li L, Sui T, Xiao L, et al. MetaKernel: Enabling Efficient Encrypted Neural Network Inference through Unified MVM and Convolution. Proceedings of the ACM on Programming Languages. 2025;9(OOPSLA2):1261-1288. doi:10.1145/3763095
BibTeX
@article{Yuan2025,
author = {Yuan, Peng and Liu, Yan and Lai, JianXin and Li, Long and Sui, Tianxiang and Xiao, Linjie and Zhang, Xiaojing and Zhu, Qing and Xue, Jingling},
title = {MetaKernel: Enabling Efficient Encrypted Neural Network Inference through Unified MVM and Convolution},
journal = {Proceedings of the ACM on Programming Languages},
year = {2025},
volume = {9},
number = {OOPSLA2},
pages = {1261--1288},
doi = {10.1145/3763095}
}
EUVIMED