3D adversarial objects generation for wider-view face recognition attacks

Lingzhuang Meng

doi:10.1016/j.knosys.2026.116111

3D adversarial objects generation for wider-view face recognition attacks

2026-05-05

0

PRE

AI

DOI：10.1016/j.knosys.2026.116111

Original

Original request for help

Save

Abstract

En 中文

• We propose a novel 3DAO for attacking face recognition in multiple views. • A Gaussian localization strategy is designed to initialize adversarial object. • Three types of losses are devised to optimize the quality 3D adversarial objects.

Keywords:

3DAO

face recognition

adversarial objects

Gaussian localization

multi-view attacks

3D adversarial objects generation for wider-view face recognition attacks

Abstract

Journal

Organization