AI Paper Atlas

Research map / Federated learning and privacy / Privacy attacks and secure learning

People Taking Photos That Faces Never Share: Privacy Protection and Fairness Enhancement from Camera to User

Junjie Zhu, Lin Gu, Xiaoxiao Wu, Zheng Li, Tatsuya Harada, Yingying Zhu

AAAI 2023 · Federated learning and privacy › Privacy attacks and secure learning · approach: Visual privacy and face de-identification · 8 citations (OpenAlex)

Read the paper

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
encrypt faces as latent variables or decode them into fake facesFFEM · LFWTop-1 face identification accuracy: 93.4% -> 0%Improved
encrypt faces as latent variablesFFEM · LFWTop-1 face identification accuracy: 96.0% -> 0.4%Improved
replace faces with fake facesFFEM · LFWTop-1 face identification accuracy: 89.4% -> 0%Improved
use encrypted latent-variable faces instead of original facesFFEM · HMDB51Top-1 action recognition accuracy: 61.70% -> 61.70%; 83.27% -> 83.27%; 82.35% -> 82.29%No clear effect
replace faces with fake faces instead of original facesFFEM · HMDB51Top-1 action recognition accuracy: 61.70% -> 61.31%; 83.27% -> 82.35%; 82.35% -> 81.70%Hurt
replace original faces with fake facesFFEM · UTK-face gender groupsMean Difference: -7.31% -> -2.60%Improved
replace original faces with fake facesFFEM · UTK-face race groupsMean Difference: -3.63% -> 1.19%Improved
replace original faces with fake facesFFEM · UTK-face gender groupsMean Difference in attractiveness ratings: 14.3% -> 7.73%Improved
increase temperature λ from 0.5 to 0.6 (ablation)FFEM · UTK-face gender groupsMean Difference: 7.73% -> 10.8%Hurt
increase temperature λ from 0.5 to 0.7 (ablation)FFEM · UTK-face gender groupsMean Difference: 7.73% -> 12.7%Hurt
increase temperature λ from 0.5 to 0.8 (ablation)FFEM · UTK-face gender groupsMean Difference: 7.73% -> 12.51%Hurt
decrease temperature λ (ablation)FFEM · UTK-face gender groupsSTD of attractiveness scores: 12.37 -> 10.25Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy