notesum.ai

Published at October 31

Adversarial Attacks of Vision Tasks in the Past 10 Years: A Survey

cs.CV

cs.CR

Released Date: October 31, 2024

Authors: Chiyu Zhang¹, Xiaogang Xu², Jiafei Wu³, Zhe Liu⁴, Lu Zhou¹

Aff.: ¹Nanjing University of Aeronautics and Astronautics, China; ²The Chinese University of Hong Kong, China; ³The University of Hong Kong, China; ⁴Zhejiang Lab, China

Arxiv: http://arxiv.org/abs/2410.23687v1

Model	Class	Backbone
ND Models	N-Basic	-
FC Models	N-Basic	FC
Normal CNNs	N-Basic	CNN
Normal ViTs	N-Basic	ViT
CLIP (Radford et al., 2021)	N-Basic	Transformer
VAE (Kingma, 2013)	N-Archi	FC, CNN
VAE-GAN (Larsen et al., 2016)	N-Archi	FC, CNN
Single ATs (Kurakin et al., 2016; Tramèr et al., 2017a)	AT	CNN
Ensemble ATs (Tramèr et al., 2017a)	AT	CNN
ARS (Salman et al., 2019)	AT	CNN
RS (Cohen et al., 2019)	AT	CNN
ALP (Kannan et al., 2018)	AT	CNN
Bit-Red (Xu, 2017)	DD-Trans	CNN
JPEG (Guo et al., 2017)	DM-Trans	CNN
R&P (Xie et al., 2017a)	DM-Trans	CNN
FD (Liu et al., 2019)	DM-Trans	CNN
NRP (Naseer et al., 2020)	DM-Deno	CNN
HGD (Liao et al., 2018)	DM-Deno	CNN
ComDefend (Jia et al., 2019)	DM-Deno	CNN
DiffPure (Nie et al., 2022)	DM-Deno	CNN, ViT
DeM (Xie et al., 2019a)	DM-Deno + AT	CNN
NIPS-r3 (3), 2017)	DM-Trans + AT	CNN