Just One Moment: Structural Vulnerability of Deep Action Recognition against One Frame Attack

Jaehui Hwang; Jun Hyuk Kim; Jun Ho Choi; Jong Seok Lee

doi:10.1109/ICCV48922.2021.00757

Just One Moment: Structural Vulnerability of Deep Action Recognition against One Frame Attack

Jaehui Hwang, Jun Hyuk Kim, Jun Ho Choi, Jong Seok Lee

School of Integrated Technology

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

15 Citations (Scopus)

Abstract

The video-based action recognition task has been extensively studied in recent years. In this paper, we study the structural vulnerability of deep learning-based action recognition models against the adversarial attack using the one frame attack that adds an inconspicuous perturbation to only a single frame of a given video clip. Our analysis shows that the models are highly vulnerable against the one frame attack due to their structural properties. Experiments demonstrate high fooling rates and inconspicuous characteristics of the attack. Furthermore, we show that strong universal one frame perturbations can be obtained under various scenarios. Our work raises the serious issue of adversarial vulnerability of the state-of-the-art action recognition models in various perspectives.

Original language	English
Title of host publication	Proceedings - 2021 IEEE/CVF International Conference on Computer Vision, ICCV 2021
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	7648-7656
Number of pages	9
ISBN (Electronic)	9781665428125
DOIs	https://doi.org/10.1109/ICCV48922.2021.00757
Publication status	Published - 2021
Event	18th IEEE/CVF International Conference on Computer Vision, ICCV 2021 - Virtual, Online, Canada Duration: 2021 Oct 11 → 2021 Oct 17

Publication series

Name	Proceedings of the IEEE International Conference on Computer Vision
ISSN (Print)	1550-5499

Conference

Conference	18th IEEE/CVF International Conference on Computer Vision, ICCV 2021
Country/Territory	Canada
City	Virtual, Online
Period	21/10/11 → 21/10/17

Bibliographical note

Publisher Copyright:
© 2021 IEEE

All Science Journal Classification (ASJC) codes

Software
Computer Vision and Pattern Recognition

Access to Document

10.1109/ICCV48922.2021.00757

Cite this

Hwang, J., Kim, J. H., Choi, J. H., & Lee, J. S. (2021). Just One Moment: Structural Vulnerability of Deep Action Recognition against One Frame Attack. In Proceedings - 2021 IEEE/CVF International Conference on Computer Vision, ICCV 2021 (pp. 7648-7656). (Proceedings of the IEEE International Conference on Computer Vision). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ICCV48922.2021.00757

Hwang, Jaehui ; Kim, Jun Hyuk ; Choi, Jun Ho et al. / Just One Moment : Structural Vulnerability of Deep Action Recognition against One Frame Attack. Proceedings - 2021 IEEE/CVF International Conference on Computer Vision, ICCV 2021. Institute of Electrical and Electronics Engineers Inc., 2021. pp. 7648-7656 (Proceedings of the IEEE International Conference on Computer Vision).

@inproceedings{9dc234cf7f9746698ae5a0fabc595651,

title = "Just One Moment: Structural Vulnerability of Deep Action Recognition against One Frame Attack",

abstract = "The video-based action recognition task has been extensively studied in recent years. In this paper, we study the structural vulnerability of deep learning-based action recognition models against the adversarial attack using the one frame attack that adds an inconspicuous perturbation to only a single frame of a given video clip. Our analysis shows that the models are highly vulnerable against the one frame attack due to their structural properties. Experiments demonstrate high fooling rates and inconspicuous characteristics of the attack. Furthermore, we show that strong universal one frame perturbations can be obtained under various scenarios. Our work raises the serious issue of adversarial vulnerability of the state-of-the-art action recognition models in various perspectives.",

author = "Jaehui Hwang and Kim, {Jun Hyuk} and Choi, {Jun Ho} and Lee, {Jong Seok}",

note = "Publisher Copyright: {\textcopyright} 2021 IEEE; 18th IEEE/CVF International Conference on Computer Vision, ICCV 2021 ; Conference date: 11-10-2021 Through 17-10-2021",

year = "2021",

doi = "10.1109/ICCV48922.2021.00757",

language = "English",

series = "Proceedings of the IEEE International Conference on Computer Vision",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "7648--7656",

booktitle = "Proceedings - 2021 IEEE/CVF International Conference on Computer Vision, ICCV 2021",

address = "United States",

}

Hwang, J, Kim, JH, Choi, JH & Lee, JS 2021, Just One Moment: Structural Vulnerability of Deep Action Recognition against One Frame Attack. in Proceedings - 2021 IEEE/CVF International Conference on Computer Vision, ICCV 2021. Proceedings of the IEEE International Conference on Computer Vision, Institute of Electrical and Electronics Engineers Inc., pp. 7648-7656, 18th IEEE/CVF International Conference on Computer Vision, ICCV 2021, Virtual, Online, Canada, 21/10/11. https://doi.org/10.1109/ICCV48922.2021.00757

Just One Moment: Structural Vulnerability of Deep Action Recognition against One Frame Attack. / Hwang, Jaehui; Kim, Jun Hyuk; Choi, Jun Ho et al.
Proceedings - 2021 IEEE/CVF International Conference on Computer Vision, ICCV 2021. Institute of Electrical and Electronics Engineers Inc., 2021. p. 7648-7656 (Proceedings of the IEEE International Conference on Computer Vision).

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

TY - GEN

T1 - Just One Moment

T2 - 18th IEEE/CVF International Conference on Computer Vision, ICCV 2021

AU - Hwang, Jaehui

AU - Kim, Jun Hyuk

AU - Choi, Jun Ho

AU - Lee, Jong Seok

PY - 2021

Y1 - 2021

N2 - The video-based action recognition task has been extensively studied in recent years. In this paper, we study the structural vulnerability of deep learning-based action recognition models against the adversarial attack using the one frame attack that adds an inconspicuous perturbation to only a single frame of a given video clip. Our analysis shows that the models are highly vulnerable against the one frame attack due to their structural properties. Experiments demonstrate high fooling rates and inconspicuous characteristics of the attack. Furthermore, we show that strong universal one frame perturbations can be obtained under various scenarios. Our work raises the serious issue of adversarial vulnerability of the state-of-the-art action recognition models in various perspectives.

AB - The video-based action recognition task has been extensively studied in recent years. In this paper, we study the structural vulnerability of deep learning-based action recognition models against the adversarial attack using the one frame attack that adds an inconspicuous perturbation to only a single frame of a given video clip. Our analysis shows that the models are highly vulnerable against the one frame attack due to their structural properties. Experiments demonstrate high fooling rates and inconspicuous characteristics of the attack. Furthermore, we show that strong universal one frame perturbations can be obtained under various scenarios. Our work raises the serious issue of adversarial vulnerability of the state-of-the-art action recognition models in various perspectives.

UR - http://www.scopus.com/inward/record.url?scp=85121658292&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85121658292&partnerID=8YFLogxK

U2 - 10.1109/ICCV48922.2021.00757

DO - 10.1109/ICCV48922.2021.00757

M3 - Conference contribution

AN - SCOPUS:85121658292

T3 - Proceedings of the IEEE International Conference on Computer Vision

SP - 7648

EP - 7656

BT - Proceedings - 2021 IEEE/CVF International Conference on Computer Vision, ICCV 2021

PB - Institute of Electrical and Electronics Engineers Inc.

Y2 - 11 October 2021 through 17 October 2021

ER -

Hwang J, Kim JH, Choi JH, Lee JS. Just One Moment: Structural Vulnerability of Deep Action Recognition against One Frame Attack. In Proceedings - 2021 IEEE/CVF International Conference on Computer Vision, ICCV 2021. Institute of Electrical and Electronics Engineers Inc. 2021. p. 7648-7656. (Proceedings of the IEEE International Conference on Computer Vision). doi: 10.1109/ICCV48922.2021.00757

Just One Moment: Structural Vulnerability of Deep Action Recognition against One Frame Attack

Abstract

Publication series

Conference

Bibliographical note

All Science Journal Classification (ASJC) codes

Access to Document

Other files and links

Fingerprint

Cite this