LaVAN: Localized and visible adversarial noise

Danny Karmon; Daniel Zoran; Yoav Goldberg

LaVAN: Localized and visible adversarial noise

Danny Karmon, Daniel Zoran, Yoav Goldberg

Department of Computer Science

Bar-Ilan University

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

Most works on adversarial examples for deeplearning based image classifiers use noise that, while small, covers the entire image. We explore the case where the noise is allowed to be visible but confined to a small, localized patch of the image, without covering any of the main object(s) in the image. We show that it is possible to generate localized adversarial noises that cover only 2% of the pixels in the image, none of them over the main object, and that are transferable across images and locations, and successfully fool a stateof-the-art Inception v3 model with very high success rates.

Original language	American English
Title of host publication	35th International Conference on Machine Learning, ICML 2018
Editors	Jennifer Dy, Andreas Krause
Pages	3903-3911
Number of pages	9
ISBN (Electronic)	9781510867963
State	Published - 1 Jan 2018
Event	35th International Conference on Machine Learning, ICML 2018 - Stockholm, Sweden Duration: 10 Jul 2018 → 15 Jul 2018

Publication series

Name	35th International Conference on Machine Learning, ICML 2018
Volume	6

Conference

Conference	35th International Conference on Machine Learning, ICML 2018
Country/Territory	Sweden
City	Stockholm
Period	10/07/18 → 15/07/18

All Science Journal Classification (ASJC) codes

Computational Theory and Mathematics
Human-Computer Interaction
Software

Cite this

@inproceedings{c59ddb8e42ba43ecab5a1aedd2d60d18,

title = "LaVAN: Localized and visible adversarial noise",

abstract = "Most works on adversarial examples for deeplearning based image classifiers use noise that, while small, covers the entire image. We explore the case where the noise is allowed to be visible but confined to a small, localized patch of the image, without covering any of the main object(s) in the image. We show that it is possible to generate localized adversarial noises that cover only 2\% of the pixels in the image, none of them over the main object, and that are transferable across images and locations, and successfully fool a stateof-the-art Inception v3 model with very high success rates.",

author = "Danny Karmon and Daniel Zoran and Yoav Goldberg",

year = "2018",

month = jan,

day = "1",

language = "American English",

series = "35th International Conference on Machine Learning, ICML 2018",

pages = "3903--3911",

editor = "Jennifer Dy and Andreas Krause",

booktitle = "35th International Conference on Machine Learning, ICML 2018",

}

LaVAN: Localized and visible adversarial noise. / Karmon, Danny; Zoran, Daniel; Goldberg, Yoav.
35th International Conference on Machine Learning, ICML 2018. ed. / Jennifer Dy; Andreas Krause. 2018. p. 3903-3911 (35th International Conference on Machine Learning, ICML 2018; Vol. 6).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - LaVAN

T2 - 35th International Conference on Machine Learning, ICML 2018

AU - Karmon, Danny

AU - Zoran, Daniel

AU - Goldberg, Yoav

PY - 2018/1/1

Y1 - 2018/1/1

N2 - Most works on adversarial examples for deeplearning based image classifiers use noise that, while small, covers the entire image. We explore the case where the noise is allowed to be visible but confined to a small, localized patch of the image, without covering any of the main object(s) in the image. We show that it is possible to generate localized adversarial noises that cover only 2% of the pixels in the image, none of them over the main object, and that are transferable across images and locations, and successfully fool a stateof-the-art Inception v3 model with very high success rates.

AB - Most works on adversarial examples for deeplearning based image classifiers use noise that, while small, covers the entire image. We explore the case where the noise is allowed to be visible but confined to a small, localized patch of the image, without covering any of the main object(s) in the image. We show that it is possible to generate localized adversarial noises that cover only 2% of the pixels in the image, none of them over the main object, and that are transferable across images and locations, and successfully fool a stateof-the-art Inception v3 model with very high success rates.

UR - http://www.scopus.com/inward/record.url?scp=85057290274&partnerID=8YFLogxK

M3 - Conference contribution

T3 - 35th International Conference on Machine Learning, ICML 2018

SP - 3903

EP - 3911

BT - 35th International Conference on Machine Learning, ICML 2018

A2 - Dy, Jennifer

A2 - Krause, Andreas

Y2 - 10 July 2018 through 15 July 2018

ER -

LaVAN: Localized and visible adversarial noise

Abstract

Publication series

Conference

All Science Journal Classification (ASJC) codes

Other files and links

Fingerprint

Cite this