arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2006.01342cs.CReess.IV
收藏

一种基于GAN的隐私保护深度神经网络图像变换方案

A GAN-Based Image Transformation Scheme for Privacy-Preserving Deep Neural Networks

Warit Sirichotedumrong, Hitoshi Kiya

摘要:

我们提出了一种新的图像变换方案,使用生成对抗网络(GANs)的隐私保护深度神经网络(DNN)。该方案不仅使我们能够将没有视觉信息的图像应用于DNN,而且还增强了对仅密文攻击(COA)(包括基于DNN的攻击)的鲁棒性。在本文中,所提出的变换方案被证明能够保护普通图像上的视觉信息,并且视觉保护的图像被直接应用于DNN用于隐私保护图像分类。由于所提出的方案利用了GANs,因此不需要管理加密密钥。在一个图像分类实验中,我们评估所提出的计划的有效性方面的分类精度和对COA的鲁棒性。

英文摘要:

We propose a novel image transformation scheme using generative adversarial networks (GANs) for privacy-preserving deep neural networks (DNNs). The proposed scheme enables us not only to apply images without visual information to DNNs, but also to enhance robustness against ciphertext-only attacks (COAs) including DNN-based attacks. In this paper, the proposed transformation scheme is demonstrated to be able to protect visual information on plain images, and the visually-protected images are directly applied to DNNs for privacy-preserving image classification. Since the proposed scheme utilizes GANs, there is no need to manage encryption keys. In an image classification experiment, we evaluate the effectiveness of the proposed scheme in terms of classification accuracy and robustness against COAs.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑