Transferable perturbations of deep feature distributions

N Inkawhich; KJ Liang; L Carin

Transferable perturbations of deep feature distributions

N Inkawhich, KJ Liang, L Carin

Research output: Contribution to journal › Article › peer-review

Abstract

Almost all current adversarial attacks of CNN classifiers rely on information derived from the output layer of the network. This work presents a new adversarial attack based on the modeling and exploitation of class-wise and layer-wise deep feature distributions. We …

Original language	English (US)
Journal	arXiv preprint arXiv:2004.12519
State	Published - 2020
Externally published	Yes

Bibliographical note

Cited By (since 2020): 8

M1 - Query date: 2021-03-11 11:12:31

M1 - 8 cites: https://scholar.google.com/scholar?cites=16609346203945225693&as_sdt=2005&sciodt=0,5&hl=en

Access to Document

https://arxiv.org/abs/2004.12519

Cite this

@article{5121f0b096b045bdb08c9a8fa0cabd8e,

title = "Transferable perturbations of deep feature distributions",

abstract = "Almost all current adversarial attacks of CNN classifiers rely on information derived from the output layer of the network. This work presents a new adversarial attack based on the modeling and exploitation of class-wise and layer-wise deep feature distributions. We …",

author = "N Inkawhich and KJ Liang and L Carin",

note = "Cited By (since 2020): 8 M1 - Query date: 2021-03-11 11:12:31 M1 - 8 cites: https://scholar.google.com/scholar?cites=16609346203945225693&as_sdt=2005&sciodt=0,5&hl=en",

year = "2020",

language = "English (US)",

journal = "arXiv preprint arXiv:2004.12519",

publisher = "arxiv.org",

}

TY - JOUR

T1 - Transferable perturbations of deep feature distributions

AU - Inkawhich, N

AU - Liang, KJ

AU - Carin, L

N1 - Cited By (since 2020): 8 M1 - Query date: 2021-03-11 11:12:31 M1 - 8 cites: https://scholar.google.com/scholar?cites=16609346203945225693&as_sdt=2005&sciodt=0,5&hl=en

PY - 2020

Y1 - 2020

N2 - Almost all current adversarial attacks of CNN classifiers rely on information derived from the output layer of the network. This work presents a new adversarial attack based on the modeling and exploitation of class-wise and layer-wise deep feature distributions. We …

AB - Almost all current adversarial attacks of CNN classifiers rely on information derived from the output layer of the network. This work presents a new adversarial attack based on the modeling and exploitation of class-wise and layer-wise deep feature distributions. We …

M3 - Article

JO - arXiv preprint arXiv:2004.12519

JF - arXiv preprint arXiv:2004.12519

ER -

Transferable perturbations of deep feature distributions

Abstract

Bibliographical note

Access to Document

Fingerprint

Cite this