Stealthy backdoor attack method targeting group fairness in self-supervised learning

Fengrui Hao

doi:10.1016/j.patcog.2026.113499

Stealthy backdoor attack method targeting group fairness in self-supervised learning

2026-03-17

0

PRE

AI

DOI：10.1016/j.patcog.2026.113499

Original

Original request for help

Save

Abstract

En 中文

• We reveal fairness vulnerability in self-supervised learning under backdoor attacks. • We propose SFBA, a stealthy backdoor attack targeting group fairness in SSL. • Experiments show SFBA effectively amplifies unfairness across classes in SSL models.

Keywords:

backdoor attack

self-supervised learning

group fairness

unfairness amplification

stealthy attack

Stealthy backdoor attack method targeting group fairness in self-supervised learning

Abstract

Journal

Organization