BriefGPT.xyz
Nov, 2020
实例编码是否能够实现私有学习？
An Attack on InstaHide: Is Private Learning Possible with Instance Encoding?
HTML
PDF
Nicholas Carlini, Samuel Deng, Sanjam Garg, Somesh Jha, Saeed Mahloujifar...
TL;DR
本研究探讨了通过实例编码机制来保护隐私的非隐私学习算法，提出了两种攻击模型，分别证明了实现（强）模型的不可能性结果和在 InstaHide 上实施较弱的攻击模型的可行性。
Abstract
A learning algorithm is private if the produced model does not reveal (too much) about its training set.
instahide
[Huang, Song, Li, Arora, ICML'20] is a recent proposal that claims to preserve
privacy
by an enco
→