Don't FREAK Out: A Frequency-Inspired Approach to Detecting Backdoor Poisoned Samples in DNNs

Hasan Abed Al Kader Hammoud; Adel Bibi; Philip H.S. Torr; Bernard Ghanem

doi:10.1109/CVPRW59228.2023.00230

Don't FREAK Out: A Frequency-Inspired Approach to Detecting Backdoor Poisoned Samples in DNNs

Hasan Abed Al Kader Hammoud^*, Adel Bibi, Philip H.S. Torr, Bernard Ghanem

^*Corresponding author for this work

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

2 Scopus citations

Abstract

In this paper we investigate the frequency sensitivity of Deep Neural Networks (DNNs) when presented with clean samples versus poisoned samples. Our analysis shows significant disparities in frequency sensitivity between these two types of samples. Building on these findings, we propose FREAK, a frequency-based poisoned sample detection algorithm that is simple yet effective. Our experimental results demonstrate the efficacy of FREAK not only against frequency backdoor attacks but also against some spatial attacks. Our work is just the first step in leveraging these insights. We believe that our analysis and proposed defense mechanism will provide a foundation for future research and development of backdoor defenses.

Original language	English (US)
Title of host publication	Proceedings - 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2023
Publisher	IEEE Computer Society
Pages	2338-2345
Number of pages	8
ISBN (Electronic)	9798350302493
DOIs	https://doi.org/10.1109/CVPRW59228.2023.00230
State	Published - 2023
Event	2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2023 - Vancouver, Canada Duration: Jun 18 2023 → Jun 22 2023

Publication series

Name	IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops
Volume	2023-June
ISSN (Print)	2160-7508
ISSN (Electronic)	2160-7516

Conference

Conference	2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2023
Country/Territory	Canada
City	Vancouver
Period	06/18/23 → 06/22/23

Bibliographical note

Funding Information:
This work was supported by the King Abdullah University of Science and Technology (KAUST) Office of Sponsored Research through the Visual Computing Center (VCC) funding, the SDAIA-KAUST Center of Excellence in Data Science and Artificial Intelligence (SDAIA-KAUST AI), and UKRI grant: Turing AI Fellowship EP/W002981/1. We also thank the Royal Academy of Engineering and FiveAI for their support. Adel Bibi has received funding from the Amazon Research Awards.

Publisher Copyright:
© 2023 IEEE.

ASJC Scopus subject areas

Computer Vision and Pattern Recognition
Electrical and Electronic Engineering

Access to Document

10.1109/CVPRW59228.2023.00230

Cite this

Al Kader Hammoud, H. A., Bibi, A., Torr, P. H. S., & Ghanem, B. (2023). Don't FREAK Out: A Frequency-Inspired Approach to Detecting Backdoor Poisoned Samples in DNNs. In Proceedings - 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2023 (pp. 2338-2345). (IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops; Vol. 2023-June). IEEE Computer Society. https://doi.org/10.1109/CVPRW59228.2023.00230

Al Kader Hammoud, Hasan Abed ; Bibi, Adel ; Torr, Philip H.S. et al. / Don't FREAK Out : A Frequency-Inspired Approach to Detecting Backdoor Poisoned Samples in DNNs. Proceedings - 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2023. IEEE Computer Society, 2023. pp. 2338-2345 (IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops).

@inproceedings{8d64cd1c384a4c48a44389e51cf2a2cd,

title = "Don't FREAK Out: A Frequency-Inspired Approach to Detecting Backdoor Poisoned Samples in DNNs",

abstract = "In this paper we investigate the frequency sensitivity of Deep Neural Networks (DNNs) when presented with clean samples versus poisoned samples. Our analysis shows significant disparities in frequency sensitivity between these two types of samples. Building on these findings, we propose FREAK, a frequency-based poisoned sample detection algorithm that is simple yet effective. Our experimental results demonstrate the efficacy of FREAK not only against frequency backdoor attacks but also against some spatial attacks. Our work is just the first step in leveraging these insights. We believe that our analysis and proposed defense mechanism will provide a foundation for future research and development of backdoor defenses.",

author = "{Al Kader Hammoud}, {Hasan Abed} and Adel Bibi and Torr, {Philip H.S.} and Bernard Ghanem",

note = "Funding Information: This work was supported by the King Abdullah University of Science and Technology (KAUST) Office of Sponsored Research through the Visual Computing Center (VCC) funding, the SDAIA-KAUST Center of Excellence in Data Science and Artificial Intelligence (SDAIA-KAUST AI), and UKRI grant: Turing AI Fellowship EP/W002981/1. We also thank the Royal Academy of Engineering and FiveAI for their support. Adel Bibi has received funding from the Amazon Research Awards. Publisher Copyright: {\textcopyright} 2023 IEEE.; 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2023 ; Conference date: 18-06-2023 Through 22-06-2023",

year = "2023",

doi = "10.1109/CVPRW59228.2023.00230",

language = "English (US)",

series = "IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops",

publisher = "IEEE Computer Society",

pages = "2338--2345",

booktitle = "Proceedings - 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2023",

address = "United States",

}

Al Kader Hammoud, HA, Bibi, A, Torr, PHS & Ghanem, B 2023, Don't FREAK Out: A Frequency-Inspired Approach to Detecting Backdoor Poisoned Samples in DNNs. in Proceedings - 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2023. IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops, vol. 2023-June, IEEE Computer Society, pp. 2338-2345, 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2023, Vancouver, Canada, 06/18/23. https://doi.org/10.1109/CVPRW59228.2023.00230

Don't FREAK Out: A Frequency-Inspired Approach to Detecting Backdoor Poisoned Samples in DNNs. / Al Kader Hammoud, Hasan Abed; Bibi, Adel; Torr, Philip H.S. et al.
Proceedings - 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2023. IEEE Computer Society, 2023. p. 2338-2345 (IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops; Vol. 2023-June).

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

TY - GEN

T1 - Don't FREAK Out

T2 - 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2023

AU - Al Kader Hammoud, Hasan Abed

AU - Bibi, Adel

AU - Torr, Philip H.S.

AU - Ghanem, Bernard

N1 - Funding Information: This work was supported by the King Abdullah University of Science and Technology (KAUST) Office of Sponsored Research through the Visual Computing Center (VCC) funding, the SDAIA-KAUST Center of Excellence in Data Science and Artificial Intelligence (SDAIA-KAUST AI), and UKRI grant: Turing AI Fellowship EP/W002981/1. We also thank the Royal Academy of Engineering and FiveAI for their support. Adel Bibi has received funding from the Amazon Research Awards. Publisher Copyright: © 2023 IEEE.

PY - 2023

Y1 - 2023

N2 - In this paper we investigate the frequency sensitivity of Deep Neural Networks (DNNs) when presented with clean samples versus poisoned samples. Our analysis shows significant disparities in frequency sensitivity between these two types of samples. Building on these findings, we propose FREAK, a frequency-based poisoned sample detection algorithm that is simple yet effective. Our experimental results demonstrate the efficacy of FREAK not only against frequency backdoor attacks but also against some spatial attacks. Our work is just the first step in leveraging these insights. We believe that our analysis and proposed defense mechanism will provide a foundation for future research and development of backdoor defenses.

AB - In this paper we investigate the frequency sensitivity of Deep Neural Networks (DNNs) when presented with clean samples versus poisoned samples. Our analysis shows significant disparities in frequency sensitivity between these two types of samples. Building on these findings, we propose FREAK, a frequency-based poisoned sample detection algorithm that is simple yet effective. Our experimental results demonstrate the efficacy of FREAK not only against frequency backdoor attacks but also against some spatial attacks. Our work is just the first step in leveraging these insights. We believe that our analysis and proposed defense mechanism will provide a foundation for future research and development of backdoor defenses.

UR - http://www.scopus.com/inward/record.url?scp=85170825518&partnerID=8YFLogxK

U2 - 10.1109/CVPRW59228.2023.00230

DO - 10.1109/CVPRW59228.2023.00230

M3 - Conference contribution

AN - SCOPUS:85170825518

T3 - IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops

SP - 2338

EP - 2345

BT - Proceedings - 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2023

PB - IEEE Computer Society

Y2 - 18 June 2023 through 22 June 2023

ER -

Al Kader Hammoud HA, Bibi A, Torr PHS, Ghanem B. Don't FREAK Out: A Frequency-Inspired Approach to Detecting Backdoor Poisoned Samples in DNNs. In Proceedings - 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2023. IEEE Computer Society. 2023. p. 2338-2345. (IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops). doi: 10.1109/CVPRW59228.2023.00230

Don't FREAK Out: A Frequency-Inspired Approach to Detecting Backdoor Poisoned Samples in DNNs

Abstract

Publication series

Conference

Bibliographical note

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this