Adversarial purification using random encoding networks

Yuxin Gong

doi:10.1016/j.asoc.2025.113604

Adversarial purification using random encoding networks

2025-07-24

0

PRE

AI

DOI：10.1016/j.asoc.2025.113604

Original

Original request for help

Save

Abstract

En 中文

• A novel adversarial defense framework named the random encoding network is proposed. • A denoiser with adversarial sparse coding and dynamic random switching is proposed. • A classifier with diversity constraint is integrated into the defense framework. • Experiments verify the random encoding network’s superior adversarial robustness.

Keywords:

adversarial defense

random encoding network

denoiser

adversarial sparse coding

diversity constraint

Adversarial purification using random encoding networks

Abstract

Journal

Organization