DeDiM: De-identification using a diffusion model

As a countermeasure against malicious authentication in a face recognition system using a face image obtained from SNS or the like, de-identification methods based on adversarial example have been studied. However, since adversarial example directly uses the gradient information of a face recognitio...

Full description

Saved in:
Bibliographic Details
Published in2022 International Conference of the Biometrics Special Interest Group (BIOSIG) pp. 1 - 5
Main Authors Uchida, Hidetsugu, Abe, Narishige, Yamada, Shigefumi
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.09.2022
Subjects
Authentication
Biological system modeling
de-identification
diffusion model
Face recognition
Image quality
Image recognition
Shape
Visualization
Online AccessGet full text

Cover

More Information
Summary:As a countermeasure against malicious authentication in a face recognition system using a face image obtained from SNS or the like, de-identification methods based on adversarial example have been studied. However, since adversarial example directly uses the gradient information of a face recognition model, it is highly dependent on the model, and a de-identification effect and image quality are difficult to achieve for an unknown recognition model. In this study, we propose a novel deidentification method based on a diffusion model, which has high generalizability to an unknown recognition model by applying minute changes to face shapes. Experiments using LFW showed that the proposed method has a higher de-identification effect for unknown models and better image quality than a conventional method using adversarial example.
ISSN:1617-5468
DOI:10.1109/BIOSIG55365.2022.9896972