Attacking all tasks at once using adversarial examples in multi-task learning

doi:10.1016/j.neucom.2025.131503

Attacking all tasks at once using adversarial examples in multi-task learning

2025-09-13

0

PRE

AI

DOI：10.1016/j.neucom.2025.131503

Original

Original request for help

Save

Abstract

En 中文

• Existing attempts on attacking multi-task models presents inherent drawbacks. • Formulate the MTL adversarial attack as an optimization problem. • Develop DGBA to efficiently solve the multi-task attack problem.

Attacking all tasks at once using adversarial examples in multi-task learning

Abstract

Journal

Organization