Saved in:
Bibliographic Details
Main Authors: Schneider, David, Sajadmanesh, Sina, Sehwag, Vikash, Sarfraz, Saquib, Stiefelhagen, Rainer, Lyu, Lingjuan, Sharma, Vivek
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2410.17098
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910751826903040
author Schneider, David
Sajadmanesh, Sina
Sehwag, Vikash
Sarfraz, Saquib
Stiefelhagen, Rainer
Lyu, Lingjuan
Sharma, Vivek
author_facet Schneider, David
Sajadmanesh, Sina
Sehwag, Vikash
Sarfraz, Saquib
Stiefelhagen, Rainer
Lyu, Lingjuan
Sharma, Vivek
contents Privacy-preserving computer vision is an important emerging problem in machine learning and artificial intelligence. Prevalent methods tackling this problem use differential privacy (DP) or obfuscation techniques to protect the privacy of individuals. In both cases, the utility of the trained model is sacrificed heavily in this process. In this work, we present an anonymization pipeline that replaces sensitive human subjects in video datasets with synthetic avatars within context, employing a combined rendering and stable diffusion-based strategy. Additionally we propose masked differential privacy ({MaskDP}) to protect non-anonymized but privacy sensitive background information. MaskDP allows for controlling sensitive regions where differential privacy is applied, in contrast to applying DP on the entire input. This combined methodology provides strong privacy protection while minimizing the usual performance penalty of privacy preserving methods. Experiments on multiple challenging action recognition datasets demonstrate that our proposed techniques result in better utility-privacy trade-offs compared to standard differentially private training in the especially demanding $ε<1$ regime.
format Preprint
id arxiv_https___arxiv_org_abs_2410_17098
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Activity Recognition on Avatar-Anonymized Datasets with Masked Differential Privacy
Schneider, David
Sajadmanesh, Sina
Sehwag, Vikash
Sarfraz, Saquib
Stiefelhagen, Rainer
Lyu, Lingjuan
Sharma, Vivek
Computer Vision and Pattern Recognition
68T45
I.4.m
Privacy-preserving computer vision is an important emerging problem in machine learning and artificial intelligence. Prevalent methods tackling this problem use differential privacy (DP) or obfuscation techniques to protect the privacy of individuals. In both cases, the utility of the trained model is sacrificed heavily in this process. In this work, we present an anonymization pipeline that replaces sensitive human subjects in video datasets with synthetic avatars within context, employing a combined rendering and stable diffusion-based strategy. Additionally we propose masked differential privacy ({MaskDP}) to protect non-anonymized but privacy sensitive background information. MaskDP allows for controlling sensitive regions where differential privacy is applied, in contrast to applying DP on the entire input. This combined methodology provides strong privacy protection while minimizing the usual performance penalty of privacy preserving methods. Experiments on multiple challenging action recognition datasets demonstrate that our proposed techniques result in better utility-privacy trade-offs compared to standard differentially private training in the especially demanding $ε<1$ regime.
title Activity Recognition on Avatar-Anonymized Datasets with Masked Differential Privacy
topic Computer Vision and Pattern Recognition
68T45
I.4.m
url https://arxiv.org/abs/2410.17098