Boosting Adversarial Transferability With Learnable Patch-Wise Masks

Adversarial examples have attracted widespread attention in security-critical applications because of their transferability across different models. Although many methods have been proposed to boost adversarial transferability, a gap still exists between capabilities and practical demand. In this ar...

Full description

Saved in:
Bibliographic Details
Published in:IEEE transactions on multimedia Vol. 26; pp. 3778 - 3787
Main Authors: Wei, Xingxing, Zhao, Shiji
Format: Journal Article
Language:English
Published: Piscataway IEEE 2024
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Subjects:
ISSN:1520-9210, 1941-0077
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Abstract Adversarial examples have attracted widespread attention in security-critical applications because of their transferability across different models. Although many methods have been proposed to boost adversarial transferability, a gap still exists between capabilities and practical demand. In this article, we argue that the model-specific discriminative regions are a key factor causing overfitting to the source model, and thus reducing the transferability to the target model. For that, a patch-wise mask is utilized to prune the model-specific regions when calculating adversarial perturbations. To accurately localize these regions, we present a learnable approach to automatically optimize the mask. Specifically, we simulate the target models in our framework, and adjust the patch-wise mask according to the feedback of the simulated models. To improve the efficiency, the differential evolutionary (DE) algorithm is utilized to search for patch-wise masks for a specific image. During iterative attacks, the learned masks are applied to the image to drop out the patches related to model-specific regions, thus making the gradients more generic and improving the adversarial transferability. The proposed approach is a preprocessing method and can be integrated with existing methods to further boost the transferability. Extensive experiments on the ImageNet dataset demonstrate the effectiveness of our method. We incorporate the proposed approach with existing methods to perform ensemble attacks and achieve an average success rate of 93.01% against seven advanced defense methods, which can effectively enhance the state-of-the-art transfer-based attack performance.
AbstractList Adversarial examples have attracted widespread attention in security-critical applications because of their transferability across different models. Although many methods have been proposed to boost adversarial transferability, a gap still exists between capabilities and practical demand. In this article, we argue that the model-specific discriminative regions are a key factor causing overfitting to the source model, and thus reducing the transferability to the target model. For that, a patch-wise mask is utilized to prune the model-specific regions when calculating adversarial perturbations. To accurately localize these regions, we present a learnable approach to automatically optimize the mask. Specifically, we simulate the target models in our framework, and adjust the patch-wise mask according to the feedback of the simulated models. To improve the efficiency, the differential evolutionary (DE) algorithm is utilized to search for patch-wise masks for a specific image. During iterative attacks, the learned masks are applied to the image to drop out the patches related to model-specific regions, thus making the gradients more generic and improving the adversarial transferability. The proposed approach is a preprocessing method and can be integrated with existing methods to further boost the transferability. Extensive experiments on the ImageNet dataset demonstrate the effectiveness of our method. We incorporate the proposed approach with existing methods to perform ensemble attacks and achieve an average success rate of 93.01% against seven advanced defense methods, which can effectively enhance the state-of-the-art transfer-based attack performance.
Author Wei, Xingxing
Zhao, Shiji
Author_xml – sequence: 1
  givenname: Xingxing
  orcidid: 0000-0002-0778-8377
  surname: Wei
  fullname: Wei, Xingxing
  email: xxwei@buaa.edu.cn
  organization: Institute of Artificial Intelligence, Beihang University, Beijing, China
– sequence: 2
  givenname: Shiji
  orcidid: 0000-0001-6033-6049
  surname: Zhao
  fullname: Zhao, Shiji
  email: zhaoshiji123@buaa.edu.cn
  organization: Institute of Artificial Intelligence, Beihang University, Beijing, China
BookMark eNp9kD1PwzAURS0EEqWwMzBEYk55tmunHkv5lFrBUNTRenFs6hKSYrtI_fekagfEwPTecM_V1Tkjx03bWEIuKQwoBXUzn80GDBgfcE6FEHBEelQNaQ5QFMfdLxjkilE4JWcxrgDoUEDRI3e3bRuTb96zcfVtQ8Tgsc7mAZvobMDS1z5ts4VPy2xqMTRY1jZ7xWSW-cJHm80wfsRzcuKwjvbicPvk7eF-PnnKpy-Pz5PxNDdMsZRLgW5Ulk5RxwrHsRxVVllnJXPGgVFGDE3BnShV2a0FcJUBR5UxAqWrWMX75Hrfuw7t18bGpFftpttUR80UByUlcNal5D5lQhtjsE4bnzD5tkkBfa0p6J0x3RnTO2P6YKwD4Q-4Dv4Tw_Y_5GqPeGvtrzgTVILkP_QmeVk
CODEN ITMUF8
CitedBy_id crossref_primary_10_1109_TIP_2025_3586485
crossref_primary_10_1016_j_imavis_2025_105722
crossref_primary_10_1007_s11263_025_02552_x
crossref_primary_10_1109_TMM_2024_3521769
Cites_doi 10.1109/CVPR.2019.00284
10.1088/1742-5468/ac9830
10.1109/CVPR.2019.00444
10.1109/TPAMI.2022.3231886
10.1109/ICCV.2015.169
10.1609/aaai.v33i01.33011028
10.1109/ICCV48922.2021.01172
10.1007/978-3-031-19772-7_34
10.1109/CVPR42600.2020.00034
10.1109/tmm.2022.3173533
10.1109/ICCV48922.2021.01585
10.1109/CVPR.2017.243
10.1201/9781351251389-8
10.1109/CVPR46437.2021.00891
10.1109/TMM.2021.3124083
10.1177/104649647100200201
10.1109/TMM.2021.3079723
10.1007/978-3-031-19772-7_32
10.1109/CVPR.2016.90
10.48550/arXiv.2010.11929
10.1109/CVPR46437.2021.00196
10.1109/CVPR52688.2022.01457
10.1109/CVPR.2018.00957
10.14722/ndss.2018.23198
10.1109/CVPR.2016.308
10.1109/TPAMI.2022.3176760
10.5555/3298023.3298188
10.1109/ICCV48922.2021.00754
10.1007/978-3-540-68830-3
10.1109/CVPR42600.2020.00124
10.1109/CVPR52688.2022.01456
10.1109/CVPR.2019.00095
10.1038/30918
10.1109/ICCV48922.2021.01204
10.1109/ICCV48922.2021.00063
10.1109/CVPR.2018.00191
10.1109/CVPR.2015.7298965
10.1109/tmm.2023.3255742
ContentType Journal Article
Copyright Copyright The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 2024
Copyright_xml – notice: Copyright The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 2024
DBID 97E
RIA
RIE
AAYXX
CITATION
7SC
7SP
8FD
JQ2
L7M
L~C
L~D
DOI 10.1109/TMM.2023.3315550
DatabaseName IEEE All-Society Periodicals Package (ASPP) 2005–Present
IEEE All-Society Periodicals Package (ASPP) 1998–Present
IEEE Electronic Library (IEL)
CrossRef
Computer and Information Systems Abstracts
Electronics & Communications Abstracts
Technology Research Database
ProQuest Computer Science Collection
Advanced Technologies Database with Aerospace
Computer and Information Systems Abstracts – Academic
Computer and Information Systems Abstracts Professional
DatabaseTitle CrossRef
Technology Research Database
Computer and Information Systems Abstracts – Academic
Electronics & Communications Abstracts
ProQuest Computer Science Collection
Computer and Information Systems Abstracts
Advanced Technologies Database with Aerospace
Computer and Information Systems Abstracts Professional
DatabaseTitleList
Technology Research Database
Database_xml – sequence: 1
  dbid: RIE
  name: IEEE/IET Electronic Library (IEL) (UW System Shared)
  url: https://ieeexplore.ieee.org/
  sourceTypes: Publisher
DeliveryMethod fulltext_linktorsrc
Discipline Engineering
Statistics
Computer Science
EISSN 1941-0077
EndPage 3787
ExternalDocumentID 10_1109_TMM_2023_3315550
10251606
Genre orig-research
GrantInformation_xml – fundername: Fundamental Research Funds for the Central Universities
  funderid: 10.13039/501100012226
– fundername: National Natural Science Foundation of China; Project of the National Natural Science Foundation of China
  grantid: 62076018
  funderid: 10.13039/501100001809
GroupedDBID -~X
0R~
29I
4.4
5GY
5VS
6IK
97E
AAJGR
AARMG
AASAJ
AAWTH
ABAZT
ABQJQ
ABVLG
ACGFO
ACGFS
ACIWK
AENEX
AETIX
AGQYO
AGSQL
AHBIQ
AI.
AIBXA
AKJIK
AKQYR
ALLEH
ALMA_UNASSIGNED_HOLDINGS
ATWAV
BEFXN
BFFAM
BGNUA
BKEBE
BPEOZ
CS3
DU5
EBS
EJD
HZ~
H~9
IFIPE
IFJZH
IPLJI
JAVBF
LAI
M43
O9-
OCL
P2P
PQQKQ
RIA
RIE
RNS
TN5
VH1
ZY4
AAYXX
CITATION
7SC
7SP
8FD
JQ2
L7M
L~C
L~D
ID FETCH-LOGICAL-c292t-65af8bbf91f27f3ab8de9efe62fcf0c9c54c73f5b9b52000fdc0f19cc5a6fd2d3
IEDL.DBID RIE
ISICitedReferencesCount 9
ISICitedReferencesURI http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=001175134300024&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
ISSN 1520-9210
IngestDate Mon Jun 30 05:42:15 EDT 2025
Sat Nov 29 03:10:12 EST 2025
Tue Nov 18 21:58:05 EST 2025
Wed Aug 27 02:11:30 EDT 2025
IsPeerReviewed true
IsScholarly true
Language English
License https://ieeexplore.ieee.org/Xplorehelp/downloads/license-information/IEEE.html
https://doi.org/10.15223/policy-029
https://doi.org/10.15223/policy-037
LinkModel DirectLink
MergedId FETCHMERGED-LOGICAL-c292t-65af8bbf91f27f3ab8de9efe62fcf0c9c54c73f5b9b52000fdc0f19cc5a6fd2d3
Notes ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ORCID 0000-0001-6033-6049
0000-0002-0778-8377
PQID 2930966032
PQPubID 75737
PageCount 10
ParticipantIDs crossref_primary_10_1109_TMM_2023_3315550
ieee_primary_10251606
crossref_citationtrail_10_1109_TMM_2023_3315550
proquest_journals_2930966032
PublicationCentury 2000
PublicationDate 20240000
2024-00-00
20240101
PublicationDateYYYYMMDD 2024-01-01
PublicationDate_xml – year: 2024
  text: 20240000
PublicationDecade 2020
PublicationPlace Piscataway
PublicationPlace_xml – name: Piscataway
PublicationTitle IEEE transactions on multimedia
PublicationTitleAbbrev TMM
PublicationYear 2024
Publisher IEEE
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Publisher_xml – name: IEEE
– name: The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
References ref12
ref14
Tramr (ref25) 2018
ref11
Goodfellow (ref5) 2015
ref17
ref16
ref19
ref18
Xie (ref33) 2018
ref50
ref46
Simonyan (ref20) 2015
ref45
ref48
ref47
ref42
ref41
ref44
ref49
ref8
ref7
Xie (ref32) 2018
ref4
ref3
ref6
ref40
Smilkov (ref21) 2017
Lin (ref13) 2019
ref35
ref34
ref37
ref36
Liu (ref15) 2017
ref31
ref30
ref2
ref39
ref38
Touvron (ref43) 2021
Krizhevsky (ref10) 2012; 25
Chakraborty (ref1) 2008; 143
Szegedy (ref24) 2014
ref23
ref26
ref22
ref28
ref27
ref29
Jia (ref9) 2020
References_xml – ident: ref34
  doi: 10.1109/CVPR.2019.00284
– ident: ref40
  doi: 10.1088/1742-5468/ac9830
– ident: ref3
  doi: 10.1109/CVPR.2019.00444
– ident: ref50
  doi: 10.1109/TPAMI.2022.3231886
– ident: ref4
  doi: 10.1109/ICCV.2015.169
– volume: 25
  start-page: 1097
  volume-title: Proc. Adv. Neural Inf. Process. Syst.
  year: 2012
  ident: ref10
  article-title: ImageNet classification with deep convolutional neural networks
– start-page: 1
  volume-title: Proc. Int. Conf. Learn. Representations
  year: 2014
  ident: ref24
  article-title: Intriguing properties of neural networks
– ident: ref14
  doi: 10.1609/aaai.v33i01.33011028
– start-page: 1
  volume-title: Proc. 6th Int. Conf. Learn. Representations
  year: 2018
  ident: ref33
  article-title: Mitigating adversarial effects through randomization
– start-page: 1
  volume-title: Proc. 3rd Int. Conf. Learn. Representations
  year: 2015
  ident: ref5
  article-title: Explaining and harnessing adversarial examples
– ident: ref42
  doi: 10.1109/ICCV48922.2021.01172
– ident: ref44
  doi: 10.1007/978-3-031-19772-7_34
– ident: ref19
  doi: 10.1109/CVPR42600.2020.00034
– ident: ref47
  doi: 10.1109/tmm.2022.3173533
– ident: ref27
  doi: 10.1109/ICCV48922.2021.01585
– ident: ref8
  doi: 10.1109/CVPR.2017.243
– start-page: 1
  volume-title: Proc. Int. Conf. Learn. Representations
  year: 2020
  ident: ref9
  article-title: Certified robustness for top-k predictions against adversarial perturbations via randomized smoothing
– ident: ref11
  doi: 10.1201/9781351251389-8
– ident: ref31
  doi: 10.1109/CVPR46437.2021.00891
– ident: ref48
  doi: 10.1109/TMM.2021.3124083
– ident: ref7
  doi: 10.1177/104649647100200201
– start-page: 1
  volume-title: Proc. 3rd IAPR Asian Conf. Pattern Recognit.
  year: 2015
  ident: ref20
  article-title: Very deep convolutional networks for large-scale image recognition
– start-page: 1
  volume-title: Proc. Int. Conf. Learn. Representations
  year: 2018
  ident: ref32
  article-title: Mitigating adversarial effects through randomization
– start-page: 10347
  volume-title: Proc. Int. Conf. Mach. Learn.
  year: 2021
  ident: ref43
  article-title: Training data-efficient image transformers & distillation through attention
– ident: ref45
  doi: 10.1109/TMM.2021.3079723
– ident: ref18
  doi: 10.1007/978-3-031-19772-7_32
– year: 2017
  ident: ref21
  article-title: Smoothgrad: Removing noise by adding noise
– ident: ref6
  doi: 10.1109/CVPR.2016.90
– ident: ref39
  doi: 10.48550/arXiv.2010.11929
– start-page: 1
  volume-title: Proc. 6th Int. Conf. Learn. Representations
  year: 2018
  ident: ref25
  article-title: Ensemble adversarial training: Attacks and defenses
– ident: ref26
  doi: 10.1109/CVPR46437.2021.00196
– ident: ref37
  doi: 10.1109/CVPR52688.2022.01457
– ident: ref2
  doi: 10.1109/CVPR.2018.00957
– ident: ref36
  doi: 10.14722/ndss.2018.23198
– ident: ref23
  doi: 10.1109/CVPR.2016.308
– ident: ref49
  doi: 10.1109/TPAMI.2022.3176760
– ident: ref22
  doi: 10.5555/3298023.3298188
– start-page: 1
  volume-title: Proc. Int. Conf. Learn. Representations
  year: 2017
  ident: ref15
  article-title: Delving into transferable adversarial examples and black-box attacks
– ident: ref28
  doi: 10.1109/ICCV48922.2021.00754
– volume: 143
  volume-title: Advances in Differential Evolution
  year: 2008
  ident: ref1
  doi: 10.1007/978-3-540-68830-3
– ident: ref30
  doi: 10.1109/CVPR42600.2020.00124
– ident: ref35
  doi: 10.1109/CVPR52688.2022.01456
– ident: ref16
  doi: 10.1109/CVPR.2019.00095
– ident: ref29
  doi: 10.1038/30918
– start-page: 1
  volume-title: Proc. Int. Conf. Learn. Representations
  year: 2019
  ident: ref13
  article-title: Nesterov accelerated gradient and scale invariance for adversarial attacks
– ident: ref41
  doi: 10.1109/ICCV48922.2021.01204
– ident: ref38
  doi: 10.1109/ICCV48922.2021.00063
– ident: ref12
  doi: 10.1109/CVPR.2018.00191
– ident: ref17
  doi: 10.1109/CVPR.2015.7298965
– ident: ref46
  doi: 10.1109/tmm.2023.3255742
SSID ssj0014507
Score 2.4516895
Snippet Adversarial examples have attracted widespread attention in security-critical applications because of their transferability across different models. Although...
SourceID proquest
crossref
ieee
SourceType Aggregation Database
Enrichment Source
Index Database
Publisher
StartPage 3778
SubjectTerms Adaptation models
Adversarial Attack
Adversarial Transferability
DNNs
Evolutionary algorithms
Iterative methods
Masks
Perturbation methods
Predictive models
Statistics
Training
Visualization
Title Boosting Adversarial Transferability With Learnable Patch-Wise Masks
URI https://ieeexplore.ieee.org/document/10251606
https://www.proquest.com/docview/2930966032
Volume 26
WOSCitedRecordID wos001175134300024&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
hasFullText 1
inHoldings 1
isFullTextHit
isPrint
journalDatabaseRights – providerCode: PRVIEE
  databaseName: IEEE/IET Electronic Library (IEL) (UW System Shared)
  customDbUrl:
  eissn: 1941-0077
  dateEnd: 99991231
  omitProxy: false
  ssIdentifier: ssj0014507
  issn: 1520-9210
  databaseCode: RIE
  dateStart: 19990101
  isFulltext: true
  titleUrlDefault: https://ieeexplore.ieee.org/
  providerName: IEEE
link http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV09T8MwED1BxVAGCqWIQkEeWBjc5jvxyFfF0qpDUbtFtmOrFahBTYvEv-fsJKgSAoktgx1Fefa988e9B3AjQi21cjn1MHOjgXAkFUiklGWuH2B-q3iYWbOJeDxO5nM2qYrVbS2MUspePlN982jP8rNcbs1WGc5wZOPICGzvx3FUFmt9HxkEoa2NRj5yKMOFTH0m6bDBdDTqG5vwvu8jfZoS-x0OsqYqPyKxpZdh658fdgxHVR5J7krgT2BPrdrQqj0aSDVl23C4IzjYhqbJLUtp5lN4vM_zwtx6JtaVueBmLBJLXlqtS_3uTzJbbhbEqrCaIisywdC9oLNlociIF69FB16GT9OHZ1qZKlDpMW9Do5DrRAjNXO3F2uciyRRTWkUeguZIJsNAxr4OBRNGkcnRmXS0y6QMeaQzL_PPoLHKV-ociNbGMBRDBqYNgfCk0EmG60PBpSsxbMguDOrfnMpKcdwYX7ylduXhsBSBSQ0waQVMF26_e7yXaht_tO0YIHbalRh0oVdDmVbzsUgxqXGMDqnvXfzS7RKa-Pag3F3pQWOz3qorOJAfiMr62g61LxaT0UY
linkProvider IEEE
linkToHtml http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV3JTsMwEB0hQKIcWEoRZfWBCwe3jrO0PrJVRZCKQxG9RbFjqxWoQU1B4u8ZOymqhEDiloOtRHn2vPEy7wGcy9Aoo72UcszcaCCZohKJlIrM8wPMb3UaZs5sojMYdEcj8VgVq7taGK21u3ymW_bRneVnuXq3W2U4w5GNIyuwvRYGAWdludb3oUEQuupoZCRGBS5lFqeSTLSHcdyyRuEt30cCtUX2SyzkbFV-xGJHML3tf37aDmxVmSS5LKHfhRU9rcP2wqWBVJO2DptLkoN1qNnsshRn3oObqzwv7L1n4nyZi9SORuLoy-hZqeD9SZ4n8zFxOqy2zIo8YvAe0-dJoUmcFi9FA556t8PrPq1sFajigs9pFKamK6URnuEd46eym2mhjY44wsaUUGGgOr4JpZBWk4mZTDHjCaXCNDIZz_x9WJ3mU30AxBhrGYpBAxOHQHIlTTfDFaJMlacwcKgmtBe_OVGV5ri1vnhN3NqDiQSBSSwwSQVMEy6-e7yVeht_tG1YIJbalRg04XgBZVLNyCLBtIZZJVKfH_7S7Qw2-sP4IXm4G9wfQQ3fFJR7LcewOp-96xNYVx-I0OzUDbsvImPUjQ
openUrl ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Ajournal&rft.genre=article&rft.atitle=Boosting+Adversarial+Transferability+With+Learnable+Patch-Wise+Masks&rft.jtitle=IEEE+transactions+on+multimedia&rft.au=Wei%2C+Xingxing&rft.au=Zhao%2C+Shiji&rft.date=2024&rft.pub=IEEE&rft.issn=1520-9210&rft.volume=26&rft.spage=3778&rft.epage=3787&rft_id=info:doi/10.1109%2FTMM.2023.3315550&rft.externalDocID=10251606
thumbnail_l http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=1520-9210&client=summon
thumbnail_m http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=1520-9210&client=summon
thumbnail_s http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=1520-9210&client=summon