paper-with-me

홈 › Papers

A Fusion-Denoising Attack on InstaHide with Data Augmentation

2021-05-17 · Xinjian Luo, Xiaokui Xiao, Yuncheng Wu, Juncheng Liu, Beng Chin Ooi

InstaHide is a state-of-the-art mechanism for protecting private training images, by mixing multiple private images and modifying them such that their visual features are indistinguishable to the naked eye. In recent work, however, Carlini et al. show that it is possible to reconstruct private images from the encrypted dataset generated by InstaHide. Nevertheless, we demonstrate that Carlini et al.'s attack can be easily defeated by incorporating data augmentation into InstaHide. This leads to a natural question: is InstaHide with data augmentation secure? In this paper, we provide a negative answer to this question, by devising an attack for recovering private images from the outputs of InstaHide even when data augmentation is present. The basic idea is to use a comparative network to identify encrypted images that are likely to correspond to the same private image, and then employ a fusion-denoising network for restoring the private image from the encrypted ones, taking into account the effects of data augmentation. Extensive experiments demonstrate the effectiveness of the proposed attack in comparison to Carlini et al.'s attack.

📄 PDF Abstract BibTeX arXiv:2105.07754

Code (1)

xj231/FDN 공식 구현 pytorch

Tasks

Data AugmentationDenoising

Similar Papers 제목 키워드 기반

DP-InstaHide: Provably Defusing Poisoning and Backdoor Attacks with Differentially Private Data Augmentations

2021-03-02 · Eitan Borgnia, Jonas Geiping, Valeriia Cherepanova, Liam Fowl 외

Data poisoning and backdoor attacks manipulate training data to induce security breaches in a victim model. These attacks can be provably deflected using differentially private (DP) training methods, although this comes …

Data Poisoning

DP-InstaHide: Data Augmentations Provably Enhance Guarantees Against Dataset Manipulations

2021-09-29 · Eitan Borgnia, Jonas Geiping, Valeriia Cherepanova, Liam H Fowl 외

Data poisoning and backdoor attacks manipulate training data to induce security breaches in a victim model. These attacks can be provably deflected using differentially private (DP) training methods, although this comes …

Data Poisoning

InstaHide's Sample Complexity When Mixing Two Private Images

2020-11-24 · Baihe Huang, Zhao Song, Runzhou Tao, Junze Yin 외

Training neural networks usually require large numbers of sensitive training data, and how to protect the privacy of training data has thus become a critical topic in deep learning research. InstaHide is a state-of-the-a…

Vocal Bursts Valence Prediction

InstaHide: Instance-hiding Schemes for Private Distributed Learning

2020-10-06 · Yangsibo Huang, Zhao Song, Kai Li, Sanjeev Arora

How can multiple distributed entities collaboratively train a shared deep net on their private data while preserving privacy? This paper introduces InstaHide, a simple encryption of training images, which can be plugged …

InstaHide’s Sample Complexity When Mixing Two Private Images

2021-09-29 · Baihe Huang, Zhao Song, Runzhou Tao, Ruizhe Zhang 외

Inspired by InstaHide challenge [Huang, Song, Li and Arora'20], [Chen, Song and Zhuo'20] recently provides one mathematical formulation of InstaHide attack problem under Gaussian images distribution. They show that it su…

Vocal Bursts Valence Prediction