Saved in:
Bibliographic Details
Main Authors: Ling, Huaming, Wang, Ying, Chen, Si, Fan, Junfeng
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2511.18976
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917100819316736
author Ling, Huaming
Wang, Ying
Chen, Si
Fan, Junfeng
author_facet Ling, Huaming
Wang, Ying
Chen, Si
Fan, Junfeng
contents We address two fundamental challenges in adapting general deep CNNs for FHE-based inference: approximating non-linear activations such as ReLU with low-degree polynomials while minimizing accuracy degradation, and overcoming the ciphertext capacity barrier that constrains high-resolution image processing on FHE inference. Our contributions are twofold: (1) a single-stage fine-tuning (SFT) strategy that directly converts pre-trained CNNs into FHE-friendly forms using low-degree polynomials, achieving competitive accuracy with minimal training overhead; and (2) a generalized interleaved packing (GIP) scheme that is compatible with feature maps of virtually arbitrary spatial resolutions, accompanied by a suite of carefully designed homomorphic operators that preserve the GIP-form encryption throughout computation. These advances enable efficient, end-to-end FHE inference across diverse CNN architectures. Experiments on CIFAR-10, ImageNet, and MS COCO demonstrate that the FHE-friendly CNNs obtained via our SFT strategy achieve accuracy comparable to baselines using ReLU or SiLU activations. Moreover, this work presents the first demonstration of FHE-based inference for YOLO architectures in object detection leveraging low-degree polynomial activations.
format Preprint
id arxiv_https___arxiv_org_abs_2511_18976
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Peregrine: One-Shot Fine-Tuning for FHE Inference of General Deep CNNs
Ling, Huaming
Wang, Ying
Chen, Si
Fan, Junfeng
Computer Vision and Pattern Recognition
We address two fundamental challenges in adapting general deep CNNs for FHE-based inference: approximating non-linear activations such as ReLU with low-degree polynomials while minimizing accuracy degradation, and overcoming the ciphertext capacity barrier that constrains high-resolution image processing on FHE inference. Our contributions are twofold: (1) a single-stage fine-tuning (SFT) strategy that directly converts pre-trained CNNs into FHE-friendly forms using low-degree polynomials, achieving competitive accuracy with minimal training overhead; and (2) a generalized interleaved packing (GIP) scheme that is compatible with feature maps of virtually arbitrary spatial resolutions, accompanied by a suite of carefully designed homomorphic operators that preserve the GIP-form encryption throughout computation. These advances enable efficient, end-to-end FHE inference across diverse CNN architectures. Experiments on CIFAR-10, ImageNet, and MS COCO demonstrate that the FHE-friendly CNNs obtained via our SFT strategy achieve accuracy comparable to baselines using ReLU or SiLU activations. Moreover, this work presents the first demonstration of FHE-based inference for YOLO architectures in object detection leveraging low-degree polynomial activations.
title Peregrine: One-Shot Fine-Tuning for FHE Inference of General Deep CNNs
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2511.18976