期刊
2021 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION WORKSHOPS, CVPRW 2021
卷 -, 期 -, 页码 3275-3283出版社
IEEE COMPUTER SOC
DOI: 10.1109/CVPRW53098.2021.00366
关键词
-
The study investigates the privacy risks associated with sharing personal photos online and proposes a novel image obfuscation framework to protect private information by maximizing uncertainty while maintaining image fidelity. The approach utilizes a discriminator for bi-directional translation and predicts an image interpolation to increase uncertainty, resulting in obfuscated images faithful to the originals with significantly higher uncertainty.
Personal photos of individuals when shared online, apart from exhibiting a myriad of memorable details, also reveals a wide range of private information and potentially entails privacy risks (e.g., online harassment, tracking). To mitigate such risks, it is crucial to study techniques that allow individuals to limit the private information leaked in visual data. We tackle this problem in a novel image obfuscation framework: to maximize entropy on inferences over targeted privacy attributes, while retaining image fidelity. We approach the problem based on an encoder-decoder style architecture, with two key novelties: (a) introducing a discriminator to perform bi-directional translation simultaneously from multiple unpaired domains; (b) predicting an image interpolation which maximizes uncertainty over a target set of attributes. We find our approach generates obfuscated images faithful to the original input images and additionally increases uncertainty by 6.2x (or up to 0.85 bits) over the non-obfuscated counterparts.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据