Adversarial detection by approximation of ensemble boundary
Despite being effective in many application areas, Deep Neural Networks (DNNs) are vulnerable to being attacked. In object recognition, the attack takes the form of a small perturbation added to an image, that causes the DNN to misclassify, but to a human appears no different. Adversarial attacks le...
Uložené v:
| Vydané v: | Neurocomputing (Amsterdam) Ročník 622; s. 129364 |
|---|---|
| Hlavný autor: | |
| Médium: | Journal Article |
| Jazyk: | English |
| Vydavateľské údaje: |
Elsevier B.V
14.03.2025
|
| Predmet: | |
| ISSN: | 0925-2312 |
| On-line prístup: | Získať plný text |
| Tagy: |
Pridať tag
Žiadne tagy, Buďte prvý, kto otaguje tento záznam!
|
| Abstract | Despite being effective in many application areas, Deep Neural Networks (DNNs) are vulnerable to being attacked. In object recognition, the attack takes the form of a small perturbation added to an image, that causes the DNN to misclassify, but to a human appears no different. Adversarial attacks lead to defences that are themselves subject to attack, and the attack/defence strategies provide important information about the properties of DNNs. In this paper, a novel method of detecting adversarial attacks is proposed for an ensemble of Deep Neural Networks (DNNs) solving two-class pattern recognition problems. The ensemble is combined using Walsh coefficients which are capable of approximating Boolean functions and thereby controlling the decision boundary complexity. The hypothesis in this paper is that decision boundaries with high curvature allow adversarial perturbations to be found, but change the curvature of the decision boundary, which is then approximated in a different way by Walsh coefficients compared to the clean images. Besides controlling boundary complexity, the coefficients also measure the correlation with class labels, which may aid in understanding the learning and transferability properties of DNNs. While the experiments here use images, the proposed approach of modelling two-class ensemble decision boundaries could in principle be applied to any application area. |
|---|---|
| AbstractList | Despite being effective in many application areas, Deep Neural Networks (DNNs) are vulnerable to being attacked. In object recognition, the attack takes the form of a small perturbation added to an image, that causes the DNN to misclassify, but to a human appears no different. Adversarial attacks lead to defences that are themselves subject to attack, and the attack/defence strategies provide important information about the properties of DNNs. In this paper, a novel method of detecting adversarial attacks is proposed for an ensemble of Deep Neural Networks (DNNs) solving two-class pattern recognition problems. The ensemble is combined using Walsh coefficients which are capable of approximating Boolean functions and thereby controlling the decision boundary complexity. The hypothesis in this paper is that decision boundaries with high curvature allow adversarial perturbations to be found, but change the curvature of the decision boundary, which is then approximated in a different way by Walsh coefficients compared to the clean images. Besides controlling boundary complexity, the coefficients also measure the correlation with class labels, which may aid in understanding the learning and transferability properties of DNNs. While the experiments here use images, the proposed approach of modelling two-class ensemble decision boundaries could in principle be applied to any application area. |
| ArticleNumber | 129364 |
| Author | Windeatt, Terry |
| Author_xml | – sequence: 1 givenname: Terry orcidid: 0000-0002-5058-9701 surname: Windeatt fullname: Windeatt, Terry email: t.windeatt@surrey.ac.uk organization: University of Surrey, CVSSP, Guildford, Surrey, GU2 7XH, UK |
| BookMark | eNp9j81KxDAcxHNYwd3VN_DQF2jNZ7tFEJZFXWHBi55DmvwDKW1Sku5i396u9expYIYZ5rdBKx88IPRAcEEwKR_bwsNZh76gmIqC0JqVfIXWuKYip4zQW7RJqcWYVHO2Rk97c4GYVHSqywyMoEcXfNZMmRqGGL5dr36NYDPwCfqmg6wJZ29UnO7QjVVdgvs_3aKv15fPwzE_fby9H_anXFMhxnyHuTDCcEwUr1UpeE1LbQU3StRCcVwJ0lSV4doyuquELomhhLESaqs0tYxtEV92dQwpRbByiPOvOEmC5RVatnKBlldouUDPteelBvO3i4Mok3bgNRgXZ0xpgvt_4AfaFmXE |
| Cites_doi | 10.1016/j.neucom.2021.10.082 10.1007/s11071-023-08456-0 10.1109/TNNLS.2018.2861579 10.1109/JPROC.2023.3238024 10.1109/TNNLS.2013.2239659 10.1109/JPROC.2021.3050042 10.1088/1361-6501/ad0f6d 10.1007/s11042-022-14021-5 10.1109/TNNLS.2021.3089134 10.58496/BJML/2023/008 10.1109/COMST.2022.3233793 10.1007/s11042-020-09167-z 10.1109/TNN.2011.2159513 10.1007/s10462-021-10125-w 10.1109/CVPR.2017.17 10.1109/TNSE.2023.3240687 10.1109/TIFS.2019.2922398 10.1145/3594869 10.1145/3398394 10.1016/j.ins.2019.05.084 10.1016/j.neucom.2023.126498 10.1007/s11063-023-11189-1 10.1109/TNN.2006.875979 10.1109/TKDE.2020.2972320 10.1109/CVPR.2016.282 10.1016/j.neucom.2023.126576 10.1080/095400996116839 10.1109/TIFS.2022.3229595 10.1109/TIFS.2021.3082327 |
| ContentType | Journal Article |
| Copyright | 2025 |
| Copyright_xml | – notice: 2025 |
| DBID | AAYXX CITATION |
| DOI | 10.1016/j.neucom.2025.129364 |
| DatabaseName | CrossRef |
| DatabaseTitle | CrossRef |
| DatabaseTitleList | |
| DeliveryMethod | fulltext_linktorsrc |
| Discipline | Computer Science |
| ExternalDocumentID | 10_1016_j_neucom_2025_129364 S0925231225000360 |
| GroupedDBID | --- --K --M .DC .~1 0R~ 123 1B1 1~. 1~5 4.4 457 4G. 53G 5VS 7-5 71M 8P~ 9JM 9JN AABNK AACTN AAEDT AAEDW AAIKJ AAKOC AALRI AAOAW AAQFI AAXKI AAXLA AAXUO AAYFN ABBOA ABCQJ ABFNM ABJNI ABMAC ACDAQ ACGFS ACRLP ACZNC ADBBV ADEZE AEBSH AEIPS AEKER AENEX AFJKZ AFKWA AFTJW AFXIZ AGHFR AGUBO AGWIK AGYEJ AHHHB AHZHX AIALX AIEXJ AIKHN AITUG AJOXV AKRWK ALMA_UNASSIGNED_HOLDINGS AMFUW AMRAJ ANKPU AOUOD AXJTR BKOJK BLXMC CS3 DU5 EBS EFJIC EO8 EO9 EP2 EP3 F5P FDB FIRID FNPLU FYGXN G-Q GBLVA GBOLZ IHE J1W KOM M41 MO0 MOBAO N9A O-L O9- OAUVE OZT P-8 P-9 P2P PC. Q38 ROL RPZ SDF SDG SDP SES SEW SPC SPCBC SSN SSV SSZ T5K ZMT ~G- 29N 9DU AAQXK AATTM AAYWO AAYXX ABWVN ABXDB ACLOT ACNNM ACRPL ACVFH ADCNI ADJOM ADMUD ADNMO AEUPX AFPUW AGQPQ AIGII AIIUN AKBMS AKYEP APXCP ASPBG AVWKF AZFZN CITATION EFKBS EFLBG EJD FEDTE FGOYB HLZ HVGLF HZ~ LG9 R2- SBC WUQ XPP ~HD |
| ID | FETCH-LOGICAL-c255t-8045d5d401a49a654926cf54da595a40751b77d4cf32875c61d21336e9fac2f33 |
| ISICitedReferencesCount | 0 |
| ISICitedReferencesURI | http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=001417399000001&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D |
| ISSN | 0925-2312 |
| IngestDate | Sat Nov 29 06:34:05 EST 2025 Sat Feb 15 15:52:13 EST 2025 |
| IsPeerReviewed | true |
| IsScholarly | true |
| Keywords | Boolean functions Adversarial robustness Deep neural networks Security Ensemble Machine learning |
| Language | English |
| LinkModel | OpenURL |
| MergedId | FETCHMERGED-LOGICAL-c255t-8045d5d401a49a654926cf54da595a40751b77d4cf32875c61d21336e9fac2f33 |
| ORCID | 0000-0002-5058-9701 |
| ParticipantIDs | crossref_primary_10_1016_j_neucom_2025_129364 elsevier_sciencedirect_doi_10_1016_j_neucom_2025_129364 |
| PublicationCentury | 2000 |
| PublicationDate | 2025-03-14 |
| PublicationDateYYYYMMDD | 2025-03-14 |
| PublicationDate_xml | – month: 03 year: 2025 text: 2025-03-14 day: 14 |
| PublicationDecade | 2020 |
| PublicationTitle | Neurocomputing (Amsterdam) |
| PublicationYear | 2025 |
| Publisher | Elsevier B.V |
| Publisher_xml | – name: Elsevier B.V |
| References | Wood, Mu, Webb, Reeve, Lujan, Brown (b19) 2023; 24 Szegedy (b29) 2014 Schmidt, Santurkar, Tsipras, Talwar, Madry (b37) 2018; 31 Lin (b31) 2021; 33 Szűcs, Kiss (b24) 2023; 82 Ortiz-Jiménez, Modas, Moosavi-Dezfooli, Frossard (b5) 2021; 109 Biggio, Corona, Maiorca, Nelson, Šrndić, Laskov, ., Roli (b30) 2013 Fawzi, Moosavi-Dezfooli, Frossard (b38) 2016; 29 Qamar, Zardari (b6) 2023 Han, Lin, Shen, Wang, Guan (b2) 2023; 55 Sen, Ravindran, Raghunathan (b53) 2020 Zhang, Benz, Lin, Karjauv, Wu, Kweon (b34) 2021 Nguyen, Fernando, Fookes, Sridharan (b4) 2023 Windeatt (b20) 2006; 17 Goodfellow, Shlens, Szegedy (b32) 2015 Ilyas, Santurkar, Tsipras, Engstrom, Tran, Madry (b35) 2019; 32 Nicolae, Sinn, Tran, Buesser, Rawat, Wistuba, ., Edwards (b43) 2023 A. Shafahi, W.R. Huang, C. Studer, S. Feizi, T. Goldstein, Are adversarial examples inevitable? Windeatt, Zor (b42) 2013; 24 . Zhang, Xie, Li, Mei, Liu (b36) 2023; 111 LeCun (b55) 2023 Hurst, Miller, Muzio (b27) 1985 Song, Wu, Song, Stojanovic (b8) 2023; 55 Addesso, Barni, Di Mauro, Matta (b15) 2021; 16 Song, Sun, Song, Stojanovic (b10) 2023; 111 Craighero, Angaroni, Stella, Damiani, Antoniotti, Graudenzi (b1) 2023; 554 Tao, Shi, Qiu, Jin, Stojanovic (b11) 2023; 35 Kurakin, Goodfellow, Bengio (b48) 2018 Verma, Swami (b54) 2019; 32 Tikhonov, Arsenin (b26) 1977 Kwon, Kim, Yoon, Choi (b51) 2021; 80 S.M. Moosavi-Dezfooli, A. Fawzi, O. Fawzi, P. Frossard, Universal adversarial perturbations, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017, pp. 1765–1773. W. He, J. Wei, X. Chen, N. Carlini, D. Song, Adversarial example defense: Ensembles of weak defenses are not strong, in: 11th USENIX Workshop on Offensive Technologies, WOOT 17, 2017. Tramer, Carlini, Brendel, Madry (b17) 2020; 33 Aldahdooh, Hamidouche, Fezza, Déforges (b50) 2022; 55 Windeatt, Zor, Camgoz (b28) 2018; 30 Krizhevsky, Sutskever, Hinton (b56) 2012; 25 Madry (b46) 2019 Serban, Poll, Visser (b49) 2020; 53 Pang, Xu, Du, Chen, Zhu (b52) 2019 He, Kim, Asghar (b3) 2023; 25 Mijwel, Esen, Shamil (b7) 2023 Zhang, Zhu, Hussain, Ye, Zhou (b14) 2023; 18 S.M. Moosavi-Dezfooli, A. Fawzi, P. Frossard, Deepfool: a simple and accurate method to fool deep neural networks, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016, pp. 2574–2582. Lian, Jia, Wu, Huang (b16) 2023; 10 Tou, Gonzales (b25) 1974 Crecchi, Melis, Sotgiu, Bacciu, Biggio (b39) 2022; 470 Deng, Mu (b22) 2024; 36 Windeatt (b18) 2008 Addesso, Cirillo, Di Mauro, Matta (b13) 2020; 15 Guo, Zhao, Li, Kuang, Zhang, Han, Tan (b23) 2019; 501 Chivukula, Yang, Liu, Zhu, Zhou (b12) 2020; 33 Song, Wu, Song, Zhang, Stojanovic (b9) 2023; 550 Tumer, Ghosh (b40) 1996; 8 Windeatt, Zor (b41) 2011; 22 Krizhevsky (b44) 2023 Biggio (10.1016/j.neucom.2025.129364_b30) 2013 Nicolae (10.1016/j.neucom.2025.129364_b43) 2023 Tramer (10.1016/j.neucom.2025.129364_b17) 2020; 33 Windeatt (10.1016/j.neucom.2025.129364_b41) 2011; 22 Fawzi (10.1016/j.neucom.2025.129364_b38) 2016; 29 Windeatt (10.1016/j.neucom.2025.129364_b20) 2006; 17 Chivukula (10.1016/j.neucom.2025.129364_b12) 2020; 33 Addesso (10.1016/j.neucom.2025.129364_b15) 2021; 16 Zhang (10.1016/j.neucom.2025.129364_b36) 2023; 111 10.1016/j.neucom.2025.129364_b21 Madry (10.1016/j.neucom.2025.129364_b46) 2019 Mijwel (10.1016/j.neucom.2025.129364_b7) 2023 10.1016/j.neucom.2025.129364_b47 Tao (10.1016/j.neucom.2025.129364_b11) 2023; 35 Lian (10.1016/j.neucom.2025.129364_b16) 2023; 10 Schmidt (10.1016/j.neucom.2025.129364_b37) 2018; 31 He (10.1016/j.neucom.2025.129364_b3) 2023; 25 Krizhevsky (10.1016/j.neucom.2025.129364_b56) 2012; 25 Han (10.1016/j.neucom.2025.129364_b2) 2023; 55 Tikhonov (10.1016/j.neucom.2025.129364_b26) 1977 Kwon (10.1016/j.neucom.2025.129364_b51) 2021; 80 Zhang (10.1016/j.neucom.2025.129364_b34) 2021 Sen (10.1016/j.neucom.2025.129364_b53) 2020 Crecchi (10.1016/j.neucom.2025.129364_b39) 2022; 470 Windeatt (10.1016/j.neucom.2025.129364_b42) 2013; 24 Guo (10.1016/j.neucom.2025.129364_b23) 2019; 501 Craighero (10.1016/j.neucom.2025.129364_b1) 2023; 554 Tumer (10.1016/j.neucom.2025.129364_b40) 1996; 8 Kurakin (10.1016/j.neucom.2025.129364_b48) 2018 Aldahdooh (10.1016/j.neucom.2025.129364_b50) 2022; 55 LeCun (10.1016/j.neucom.2025.129364_b55) 2023 Ortiz-Jiménez (10.1016/j.neucom.2025.129364_b5) 2021; 109 Wood (10.1016/j.neucom.2025.129364_b19) 2023; 24 Goodfellow (10.1016/j.neucom.2025.129364_b32) 2015 Zhang (10.1016/j.neucom.2025.129364_b14) 2023; 18 Tou (10.1016/j.neucom.2025.129364_b25) 1974 Song (10.1016/j.neucom.2025.129364_b8) 2023; 55 10.1016/j.neucom.2025.129364_b45 Lin (10.1016/j.neucom.2025.129364_b31) 2021; 33 Song (10.1016/j.neucom.2025.129364_b9) 2023; 550 Windeatt (10.1016/j.neucom.2025.129364_b18) 2008 Qamar (10.1016/j.neucom.2025.129364_b6) 2023 Szűcs (10.1016/j.neucom.2025.129364_b24) 2023; 82 Ilyas (10.1016/j.neucom.2025.129364_b35) 2019; 32 Pang (10.1016/j.neucom.2025.129364_b52) 2019 Windeatt (10.1016/j.neucom.2025.129364_b28) 2018; 30 Addesso (10.1016/j.neucom.2025.129364_b13) 2020; 15 Hurst (10.1016/j.neucom.2025.129364_b27) 1985 Serban (10.1016/j.neucom.2025.129364_b49) 2020; 53 Szegedy (10.1016/j.neucom.2025.129364_b29) 2014 10.1016/j.neucom.2025.129364_b33 Song (10.1016/j.neucom.2025.129364_b10) 2023; 111 Nguyen (10.1016/j.neucom.2025.129364_b4) 2023 Deng (10.1016/j.neucom.2025.129364_b22) 2024; 36 Krizhevsky (10.1016/j.neucom.2025.129364_b44) 2023 Verma (10.1016/j.neucom.2025.129364_b54) 2019; 32 |
| References_xml | – start-page: 42 year: 2023 end-page: 45 ident: b7 article-title: Overview of neural networks publication-title: Babylon. J. Mach. Learn. – volume: 32 year: 2019 ident: b35 article-title: Adversarial examples are not bugs, they are features publication-title: Adv. Neural Inf. Process. Syst. – volume: 111 start-page: 12181 year: 2023 end-page: 12196 ident: b10 article-title: Finite-time adaptive neural resilient DSC for fractional-order nonlinear large-scale systems against sensor-actuator faults publication-title: Nonlinear Dyn. – volume: 36 year: 2024 ident: b22 article-title: Understanding and improving ensemble adversarial defense publication-title: Adv. Neural Inf. Process. Syst. – year: 2015 ident: b32 article-title: Explaining and harnessing adversarial examples – volume: 31 year: 2018 ident: b37 article-title: Adversarially robust generalization requires more data publication-title: Adv. Neural Inf. Process. Syst. – volume: 55 start-page: 1 year: 2023 end-page: 38 ident: b2 article-title: Interpreting adversarial examples in deep learning: A review publication-title: ACM Comput. Surv. – volume: 32 year: 2019 ident: b54 article-title: Error correcting output codes improve probability estimation and adversarial robustness of deep neural networks publication-title: Adv. Neural Inf. Process. Syst. – volume: 470 start-page: 257 year: 2022 end-page: 268 ident: b39 article-title: Fader: Fast adversarial example rejection publication-title: Neurocomputing – year: 2023 ident: b4 article-title: Physical adversarial attacks for surveillance: A survey publication-title: IEEE Trans. Neural Netw. Learn. Syst. – volume: 30 start-page: 1272 year: 2018 end-page: 1277 ident: b28 article-title: Approximation of ensemble boundary using spectral coefficients publication-title: IEEE Trans. Neural Netw. Learn. Syst. – year: 2020 ident: b53 article-title: Ensembles of mixed precision deep networks for increased robustness against adversarial attacks – start-page: 133 year: 2008 end-page: 147 ident: b18 article-title: Ensemble MLP classifier design publication-title: Computational Intelligence Paradigms: Innovative Applications – reference: A. Shafahi, W.R. Huang, C. Studer, S. Feizi, T. Goldstein, Are adversarial examples inevitable?, – volume: 109 start-page: 635 year: 2021 end-page: 659 ident: b5 article-title: Optimism in the face of adversity: Understanding and improving deep learning through adversarial robustness publication-title: Proc. IEEE – volume: 10 start-page: 2086 year: 2023 end-page: 2097 ident: b16 article-title: A Stackelberg game approach to the stability of networked switched systems under DoS attacks publication-title: IEEE Trans. Netw. Sci. Eng. – year: 2023 ident: b44 article-title: CIFAR-10 dataset – year: 1977 ident: b26 article-title: Solutions of Ill-Posed Problems – volume: 55 start-page: 8997 year: 2023 end-page: 9018 ident: b8 article-title: Switching-like event-triggered state estimation for reaction–diffusion neural networks against DoS attacks publication-title: Neural Process. Lett. – start-page: 124 year: 2023 end-page: 133 ident: b6 article-title: Artificial neural networks: An overview publication-title: Mesop. J. Comput. Sci. – volume: 29 year: 2016 ident: b38 article-title: Robustness of classifiers: from adversarial to random noise publication-title: Adv. Neural Inf. Process. Syst. – volume: 22 start-page: 1334 year: 2011 end-page: 1339 ident: b41 article-title: Minimising added classification error using Walsh coefficients publication-title: IEEE Trans. Neural Netw. – year: 2019 ident: b46 article-title: Towards deep learning models resistant to adversarial attacks publication-title: Towards Deep Learning Models Resistant to Adversarial Attacks – volume: 33 start-page: 1633 year: 2020 end-page: 1645 ident: b17 article-title: On adaptive attacks to adversarial example defenses publication-title: Adv. Neural Inf. Process. Syst. – start-page: 4970 year: 2019 end-page: 4979 ident: b52 article-title: Improving adversarial robustness via promoting ensemble diversity publication-title: International Conference on Machine Learning – year: 2023 ident: b55 article-title: MNIST dataset – volume: 18 start-page: 1349 year: 2023 end-page: 1364 ident: b14 article-title: A game-theoretic method for defending against advanced persistent threats in cyber systems publication-title: IEEE Trans. Inf. Forensics Secur. – volume: 16 start-page: 3604 year: 2021 end-page: 3619 ident: b15 article-title: Adversarial Kendall’s model towards containment of distributed cyber-threats publication-title: IEEE Trans. Inf. Forensics Secur. – reference: S.M. Moosavi-Dezfooli, A. Fawzi, P. Frossard, Deepfool: a simple and accurate method to fool deep neural networks, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016, pp. 2574–2582. – volume: 17 start-page: 1194 year: 2006 end-page: 1211 ident: b20 article-title: Accuracy/diversity and ensemble MLP classifier design publication-title: IEEE Trans. Neural Netw. – volume: 550 year: 2023 ident: b9 article-title: Bipartite synchronization for cooperative-competitive neural networks with reaction–diffusion terms via dual event-triggered mechanism publication-title: Neurocomputing – year: 2023 ident: b43 article-title: Adversarial robustness toolbox – start-page: 151 year: 1974 end-page: 154 ident: b25 article-title: Pattern Recognition Principles – volume: 111 start-page: 185 year: 2023 end-page: 215 ident: b36 article-title: A survey on learning to reject publication-title: Proc. IEEE – reference: S.M. Moosavi-Dezfooli, A. Fawzi, O. Fawzi, P. Frossard, Universal adversarial perturbations, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017, pp. 1765–1773. – year: 2014 ident: b29 article-title: Intriguing properties of neural networks – volume: 33 start-page: 3568 year: 2020 end-page: 3581 ident: b12 article-title: Game theoretical adversarial deep learning with variational adversaries publication-title: IEEE Trans. Knowl. Data Eng. – volume: 8 start-page: 385 year: 1996 end-page: 404 ident: b40 article-title: Error correlation and error reduction in ensemble classifiers publication-title: Connect. Sci. – volume: 33 start-page: 7888 year: 2021 end-page: 7898 ident: b31 article-title: Supervised learning in neural networks: Feedback-network-free implementation and biological plausibility publication-title: IEEE Trans. Neural Netw. Learn. Syst. – year: 2021 ident: b34 article-title: A survey on universal adversarial attack – volume: 25 year: 2012 ident: b56 article-title: Imagenet classification with deep convolutional neural networks publication-title: Adv. Neural Inf. Process. Syst. – volume: 55 start-page: 4403 year: 2022 end-page: 4462 ident: b50 article-title: Adversarial example detection for DNN models: A review and experimental comparison publication-title: Artif. Intell. Rev. – year: 1985 ident: b27 article-title: Spectral Techniques in Digital Logic – volume: 501 start-page: 182 year: 2019 end-page: 192 ident: b23 article-title: Detecting adversarial examples via prediction difference for deep neural networks publication-title: Inform. Sci. – volume: 24 start-page: 673 year: 2013 end-page: 678 ident: b42 article-title: Ensemble pruning using spectral coefficients publication-title: IEEE Trans. Neural Netw. Learn. Syst. – volume: 24 start-page: 1 year: 2023 end-page: 49 ident: b19 article-title: A unified theory of diversity in ensemble learning publication-title: J. Mach. Learn. Res. – start-page: 387 year: 2013 end-page: 402 ident: b30 article-title: Evasion attacks against machine learning at test time publication-title: Machine Learning and Knowledge Discovery in Databases: European Conference, ECML PKDD 2013, Prague, Czech Republic, September 23-27, 2013, Proceedings, Part III 13 – volume: 35 year: 2023 ident: b11 article-title: Planetary gearbox fault diagnosis based on FDKNN-DGAT with few labeled data publication-title: Meas. Sci. Technol. – volume: 80 start-page: 10339 year: 2021 end-page: 10360 ident: b51 article-title: Classification score approach for detecting adversarial example in deep neural network publication-title: Multimedia Tools Appl. – reference: . – reference: W. He, J. Wei, X. Chen, N. Carlini, D. Song, Adversarial example defense: Ensembles of weak defenses are not strong, in: 11th USENIX Workshop on Offensive Technologies, WOOT 17, 2017. – volume: 82 start-page: 16717 year: 2023 end-page: 16740 ident: b24 article-title: 2N labeling defense method against adversarial attacks by filtering and extended class label set publication-title: Multimedia Tools Appl. – volume: 25 start-page: 538 year: 2023 end-page: 566 ident: b3 article-title: Adversarial machine learning for network intrusion detection systems: A comprehensive survey publication-title: IEEE Commun. Surv. Tutor. – volume: 554 year: 2023 ident: b1 article-title: Unity is strength: Improving the detection of adversarial examples with ensemble approaches publication-title: Neurocomputing – volume: 53 start-page: 1 year: 2020 end-page: 38 ident: b49 article-title: Adversarial examples on object recognition: A comprehensive survey publication-title: ACM Comput. Surv. – volume: 15 start-page: 943 year: 2020 end-page: 958 ident: b13 article-title: ADVoIP: Adversarial detection of encrypted and concealed VoIP publication-title: IEEE Trans. Inf. Forensics Secur. – start-page: 99 year: 2018 end-page: 112 ident: b48 article-title: Adversarial examples in the physical world publication-title: Artificial Intelligence Safety and Security – volume: 470 start-page: 257 year: 2022 ident: 10.1016/j.neucom.2025.129364_b39 article-title: Fader: Fast adversarial example rejection publication-title: Neurocomputing doi: 10.1016/j.neucom.2021.10.082 – year: 2015 ident: 10.1016/j.neucom.2025.129364_b32 – volume: 31 year: 2018 ident: 10.1016/j.neucom.2025.129364_b37 article-title: Adversarially robust generalization requires more data publication-title: Adv. Neural Inf. Process. Syst. – year: 2023 ident: 10.1016/j.neucom.2025.129364_b44 – volume: 111 start-page: 12181 issue: 13 year: 2023 ident: 10.1016/j.neucom.2025.129364_b10 article-title: Finite-time adaptive neural resilient DSC for fractional-order nonlinear large-scale systems against sensor-actuator faults publication-title: Nonlinear Dyn. doi: 10.1007/s11071-023-08456-0 – start-page: 124 issue: 2023 year: 2023 ident: 10.1016/j.neucom.2025.129364_b6 article-title: Artificial neural networks: An overview publication-title: Mesop. J. Comput. Sci. – volume: 36 year: 2024 ident: 10.1016/j.neucom.2025.129364_b22 article-title: Understanding and improving ensemble adversarial defense publication-title: Adv. Neural Inf. Process. Syst. – volume: 30 start-page: 1272 issue: 4 year: 2018 ident: 10.1016/j.neucom.2025.129364_b28 article-title: Approximation of ensemble boundary using spectral coefficients publication-title: IEEE Trans. Neural Netw. Learn. Syst. doi: 10.1109/TNNLS.2018.2861579 – volume: 111 start-page: 185 issue: 2 year: 2023 ident: 10.1016/j.neucom.2025.129364_b36 article-title: A survey on learning to reject publication-title: Proc. IEEE doi: 10.1109/JPROC.2023.3238024 – volume: 24 start-page: 673 issue: 4 year: 2013 ident: 10.1016/j.neucom.2025.129364_b42 article-title: Ensemble pruning using spectral coefficients publication-title: IEEE Trans. Neural Netw. Learn. Syst. doi: 10.1109/TNNLS.2013.2239659 – volume: 109 start-page: 635 issue: 5 year: 2021 ident: 10.1016/j.neucom.2025.129364_b5 article-title: Optimism in the face of adversity: Understanding and improving deep learning through adversarial robustness publication-title: Proc. IEEE doi: 10.1109/JPROC.2021.3050042 – volume: 35 issue: 2 year: 2023 ident: 10.1016/j.neucom.2025.129364_b11 article-title: Planetary gearbox fault diagnosis based on FDKNN-DGAT with few labeled data publication-title: Meas. Sci. Technol. doi: 10.1088/1361-6501/ad0f6d – volume: 82 start-page: 16717 issue: 11 year: 2023 ident: 10.1016/j.neucom.2025.129364_b24 article-title: 2N labeling defense method against adversarial attacks by filtering and extended class label set publication-title: Multimedia Tools Appl. doi: 10.1007/s11042-022-14021-5 – year: 1985 ident: 10.1016/j.neucom.2025.129364_b27 – volume: 33 start-page: 7888 issue: 12 year: 2021 ident: 10.1016/j.neucom.2025.129364_b31 article-title: Supervised learning in neural networks: Feedback-network-free implementation and biological plausibility publication-title: IEEE Trans. Neural Netw. Learn. Syst. doi: 10.1109/TNNLS.2021.3089134 – year: 2020 ident: 10.1016/j.neucom.2025.129364_b53 – volume: 33 start-page: 1633 year: 2020 ident: 10.1016/j.neucom.2025.129364_b17 article-title: On adaptive attacks to adversarial example defenses publication-title: Adv. Neural Inf. Process. Syst. – start-page: 42 issue: 2023 year: 2023 ident: 10.1016/j.neucom.2025.129364_b7 article-title: Overview of neural networks publication-title: Babylon. J. Mach. Learn. doi: 10.58496/BJML/2023/008 – volume: 24 start-page: 1 issue: 359 year: 2023 ident: 10.1016/j.neucom.2025.129364_b19 article-title: A unified theory of diversity in ensemble learning publication-title: J. Mach. Learn. Res. – volume: 32 year: 2019 ident: 10.1016/j.neucom.2025.129364_b35 article-title: Adversarial examples are not bugs, they are features publication-title: Adv. Neural Inf. Process. Syst. – year: 2014 ident: 10.1016/j.neucom.2025.129364_b29 – volume: 25 start-page: 538 issue: 1 year: 2023 ident: 10.1016/j.neucom.2025.129364_b3 article-title: Adversarial machine learning for network intrusion detection systems: A comprehensive survey publication-title: IEEE Commun. Surv. Tutor. doi: 10.1109/COMST.2022.3233793 – volume: 80 start-page: 10339 year: 2021 ident: 10.1016/j.neucom.2025.129364_b51 article-title: Classification score approach for detecting adversarial example in deep neural network publication-title: Multimedia Tools Appl. doi: 10.1007/s11042-020-09167-z – volume: 22 start-page: 1334 issue: 8 year: 2011 ident: 10.1016/j.neucom.2025.129364_b41 article-title: Minimising added classification error using Walsh coefficients publication-title: IEEE Trans. Neural Netw. doi: 10.1109/TNN.2011.2159513 – start-page: 133 year: 2008 ident: 10.1016/j.neucom.2025.129364_b18 article-title: Ensemble MLP classifier design – volume: 55 start-page: 4403 issue: 6 year: 2022 ident: 10.1016/j.neucom.2025.129364_b50 article-title: Adversarial example detection for DNN models: A review and experimental comparison publication-title: Artif. Intell. Rev. doi: 10.1007/s10462-021-10125-w – start-page: 387 year: 2013 ident: 10.1016/j.neucom.2025.129364_b30 article-title: Evasion attacks against machine learning at test time – year: 1977 ident: 10.1016/j.neucom.2025.129364_b26 – ident: 10.1016/j.neucom.2025.129364_b33 – ident: 10.1016/j.neucom.2025.129364_b47 doi: 10.1109/CVPR.2017.17 – volume: 10 start-page: 2086 year: 2023 ident: 10.1016/j.neucom.2025.129364_b16 article-title: A Stackelberg game approach to the stability of networked switched systems under DoS attacks publication-title: IEEE Trans. Netw. Sci. Eng. doi: 10.1109/TNSE.2023.3240687 – volume: 15 start-page: 943 year: 2020 ident: 10.1016/j.neucom.2025.129364_b13 article-title: ADVoIP: Adversarial detection of encrypted and concealed VoIP publication-title: IEEE Trans. Inf. Forensics Secur. doi: 10.1109/TIFS.2019.2922398 – volume: 55 start-page: 1 issue: 14s year: 2023 ident: 10.1016/j.neucom.2025.129364_b2 article-title: Interpreting adversarial examples in deep learning: A review publication-title: ACM Comput. Surv. doi: 10.1145/3594869 – volume: 53 start-page: 1 issue: 3 year: 2020 ident: 10.1016/j.neucom.2025.129364_b49 article-title: Adversarial examples on object recognition: A comprehensive survey publication-title: ACM Comput. Surv. doi: 10.1145/3398394 – volume: 32 year: 2019 ident: 10.1016/j.neucom.2025.129364_b54 article-title: Error correcting output codes improve probability estimation and adversarial robustness of deep neural networks publication-title: Adv. Neural Inf. Process. Syst. – volume: 501 start-page: 182 year: 2019 ident: 10.1016/j.neucom.2025.129364_b23 article-title: Detecting adversarial examples via prediction difference for deep neural networks publication-title: Inform. Sci. doi: 10.1016/j.ins.2019.05.084 – volume: 29 year: 2016 ident: 10.1016/j.neucom.2025.129364_b38 article-title: Robustness of classifiers: from adversarial to random noise publication-title: Adv. Neural Inf. Process. Syst. – volume: 550 year: 2023 ident: 10.1016/j.neucom.2025.129364_b9 article-title: Bipartite synchronization for cooperative-competitive neural networks with reaction–diffusion terms via dual event-triggered mechanism publication-title: Neurocomputing doi: 10.1016/j.neucom.2023.126498 – year: 2023 ident: 10.1016/j.neucom.2025.129364_b4 article-title: Physical adversarial attacks for surveillance: A survey publication-title: IEEE Trans. Neural Netw. Learn. Syst. – volume: 55 start-page: 8997 issue: 7 year: 2023 ident: 10.1016/j.neucom.2025.129364_b8 article-title: Switching-like event-triggered state estimation for reaction–diffusion neural networks against DoS attacks publication-title: Neural Process. Lett. doi: 10.1007/s11063-023-11189-1 – year: 2021 ident: 10.1016/j.neucom.2025.129364_b34 – year: 2019 ident: 10.1016/j.neucom.2025.129364_b46 article-title: Towards deep learning models resistant to adversarial attacks – volume: 17 start-page: 1194 issue: 5 year: 2006 ident: 10.1016/j.neucom.2025.129364_b20 article-title: Accuracy/diversity and ensemble MLP classifier design publication-title: IEEE Trans. Neural Netw. doi: 10.1109/TNN.2006.875979 – volume: 33 start-page: 3568 issue: 11 year: 2020 ident: 10.1016/j.neucom.2025.129364_b12 article-title: Game theoretical adversarial deep learning with variational adversaries publication-title: IEEE Trans. Knowl. Data Eng. doi: 10.1109/TKDE.2020.2972320 – ident: 10.1016/j.neucom.2025.129364_b45 doi: 10.1109/CVPR.2016.282 – ident: 10.1016/j.neucom.2025.129364_b21 – start-page: 151 year: 1974 ident: 10.1016/j.neucom.2025.129364_b25 – volume: 554 year: 2023 ident: 10.1016/j.neucom.2025.129364_b1 article-title: Unity is strength: Improving the detection of adversarial examples with ensemble approaches publication-title: Neurocomputing doi: 10.1016/j.neucom.2023.126576 – volume: 25 year: 2012 ident: 10.1016/j.neucom.2025.129364_b56 article-title: Imagenet classification with deep convolutional neural networks publication-title: Adv. Neural Inf. Process. Syst. – start-page: 4970 year: 2019 ident: 10.1016/j.neucom.2025.129364_b52 article-title: Improving adversarial robustness via promoting ensemble diversity – year: 2023 ident: 10.1016/j.neucom.2025.129364_b43 – start-page: 99 year: 2018 ident: 10.1016/j.neucom.2025.129364_b48 article-title: Adversarial examples in the physical world – volume: 8 start-page: 385 issue: 3–4 year: 1996 ident: 10.1016/j.neucom.2025.129364_b40 article-title: Error correlation and error reduction in ensemble classifiers publication-title: Connect. Sci. doi: 10.1080/095400996116839 – year: 2023 ident: 10.1016/j.neucom.2025.129364_b55 – volume: 18 start-page: 1349 year: 2023 ident: 10.1016/j.neucom.2025.129364_b14 article-title: A game-theoretic method for defending against advanced persistent threats in cyber systems publication-title: IEEE Trans. Inf. Forensics Secur. doi: 10.1109/TIFS.2022.3229595 – volume: 16 start-page: 3604 year: 2021 ident: 10.1016/j.neucom.2025.129364_b15 article-title: Adversarial Kendall’s model towards containment of distributed cyber-threats publication-title: IEEE Trans. Inf. Forensics Secur. doi: 10.1109/TIFS.2021.3082327 |
| SSID | ssj0017129 |
| Score | 2.4371104 |
| Snippet | Despite being effective in many application areas, Deep Neural Networks (DNNs) are vulnerable to being attacked. In object recognition, the attack takes the... |
| SourceID | crossref elsevier |
| SourceType | Index Database Publisher |
| StartPage | 129364 |
| SubjectTerms | Adversarial robustness Boolean functions Deep neural networks Ensemble Machine learning Security |
| Title | Adversarial detection by approximation of ensemble boundary |
| URI | https://dx.doi.org/10.1016/j.neucom.2025.129364 |
| Volume | 622 |
| WOSCitedRecordID | wos001417399000001&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: PRVESC databaseName: Elsevier SD Freedom Collection Journals 2021 issn: 0925-2312 databaseCode: AIEXJ dateStart: 19950101 customDbUrl: isFulltext: true dateEnd: 99991231 titleUrlDefault: https://www.sciencedirect.com omitProxy: false ssIdentifier: ssj0017129 providerName: Elsevier |
| link | http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwtV1LS8NAEF60evDiW6wv9uA1YpLdbIKnIop6EA8Vewv7Cig0ljaV-u-dfSS1Voo92EMoSzoJ-ba730xm5kPoHDYdrcK0gP83ywIiwzBIOVEBKwgxG5qiRFixCfb4mPZ62ZOX9BtZOQFWlulkkg3-FWoYA7BN6ewScDdGYQC-A-hwBNjh-CfgrcTyiFs1DqUr7bTAgWXa9uGT135DEsGD1X1TOSWsttJscbTt2iGt5oOPJnT6pqmCMjOoiR68mG6L3L1g6uqhN-GjCBE1aVSuerMOB8IYcL2ZlTFxJcN-bTPMwHUcn1t2XQTg7aLUY5ODYy5wMT19tsv1j92nyQms083ecmclN1ZyZ2UVrUWMZmkLrXXub3oPzXsiFkaum6K_-7o40mbwzd_N7-TjG6HobqNN7wngjkNwB63ochdt1Sob2C-6e-jqG6C4ARSLTzwDKH4vcA0orgHdR8-3N93ru8BLXgQSfLsK-AKhiipwejnJeGLa5yWyoERxmlEOzjcNBWOKyCIGV5fKJFRRGMeJzgouoyKOD1CrfC_1IcKC0YRICYQ7FUTqS6HhI0SYMU41-PVtFNTPIh-4zib5IgzaiNUPLPfszLGuHGbBwl8eLXmlY7QxnaInqFUNx_oUrcuP6nU0PPNT4AuGuFxe |
| linkProvider | Elsevier |
| 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=Adversarial+detection+by+approximation+of+ensemble+boundary&rft.jtitle=Neurocomputing+%28Amsterdam%29&rft.au=Windeatt%2C+Terry&rft.date=2025-03-14&rft.issn=0925-2312&rft.volume=622&rft.spage=129364&rft_id=info:doi/10.1016%2Fj.neucom.2025.129364&rft.externalDBID=n%2Fa&rft.externalDocID=10_1016_j_neucom_2025_129364 |
| thumbnail_l | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=0925-2312&client=summon |
| thumbnail_m | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=0925-2312&client=summon |
| thumbnail_s | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=0925-2312&client=summon |