BOSC: A Backdoor-Based Framework for Open Set Synthetic Image Attribution
With the continuous progress of AI technology, new generative architectures continuously appear, thus driving the attention of researchers towards the development of synthetic image attribution methods capable of working in open-set scenarios. Existing approaches focus on extracting highly discrimin...
Gespeichert in:
| Veröffentlicht in: | IEEE transactions on information forensics and security Jg. 20; S. 8043 - 8058 |
|---|---|
| Hauptverfasser: | , , |
| Format: | Journal Article |
| Sprache: | Englisch |
| Veröffentlicht: |
IEEE
2025
|
| Schlagworte: | |
| ISSN: | 1556-6013, 1556-6021 |
| Online-Zugang: | Volltext |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| Abstract | With the continuous progress of AI technology, new generative architectures continuously appear, thus driving the attention of researchers towards the development of synthetic image attribution methods capable of working in open-set scenarios. Existing approaches focus on extracting highly discriminative features for closed-set architectures, increasing the confidence of the prediction when the samples come from closed-set models/architectures, or estimating the distribution of unknown samples, i.e., samples from unknown architectures. In this paper, we propose a novel framework for open set attribution of synthetic images, named BOSC (Backdoor-based Open Set Classification), that relies on backdoor injection to design a classifier with rejection option. BOSC works by deliberately including class-specific triggers inside a portion of the images in the training set to induce the network to establish a matching between in-set class features and trigger features. The behavior of the trained model with respect to samples containing a trigger is then exploited at inference time to perform sample rejection using an ad-hoc score. Experiments show that the proposed method has good performance, always surpassing the state-of-the-art. Robustness against image processing is also very good. Although we designed our method for the task of synthetic image attribution, the proposed framework is a general one and can be used for other image forensic applications. |
|---|---|
| AbstractList | With the continuous progress of AI technology, new generative architectures continuously appear, thus driving the attention of researchers towards the development of synthetic image attribution methods capable of working in open-set scenarios. Existing approaches focus on extracting highly discriminative features for closed-set architectures, increasing the confidence of the prediction when the samples come from closed-set models/architectures, or estimating the distribution of unknown samples, i.e., samples from unknown architectures. In this paper, we propose a novel framework for open set attribution of synthetic images, named BOSC (Backdoor-based Open Set Classification), that relies on backdoor injection to design a classifier with rejection option. BOSC works by deliberately including class-specific triggers inside a portion of the images in the training set to induce the network to establish a matching between in-set class features and trigger features. The behavior of the trained model with respect to samples containing a trigger is then exploited at inference time to perform sample rejection using an ad-hoc score. Experiments show that the proposed method has good performance, always surpassing the state-of-the-art. Robustness against image processing is also very good. Although we designed our method for the task of synthetic image attribution, the proposed framework is a general one and can be used for other image forensic applications. |
| Author | Barni, Mauro Wang, Jun Tondi, Benedetta |
| Author_xml | – sequence: 1 givenname: Jun orcidid: 0000-0002-0936-7904 surname: Wang fullname: Wang, Jun email: wangjunsdnu@gmail.com organization: Department of Information Engineering and Mathematics, University of Siena, Siena, Italy – sequence: 2 givenname: Benedetta orcidid: 0000-0002-7518-046X surname: Tondi fullname: Tondi, Benedetta organization: Department of Information Engineering and Mathematics, University of Siena, Siena, Italy – sequence: 3 givenname: Mauro orcidid: 0000-0002-7368-0866 surname: Barni fullname: Barni, Mauro organization: Department of Information Engineering and Mathematics, University of Siena, Siena, Italy |
| BookMark | eNpFkM1OwkAUhScGEwF9ABMX8wLFufPTad1BI9qEpIviupkpd7QiHTIdY3h7IRBdnbM431l8EzLqfY-E3AObAbD8cV0u6xlnXM2EyrkScEXGoFSapIzD6K-DuCGTYfhkTEpIszEpF1VdPNE5XZh2u_E-JAsz4IYug9nhjw9b6nyg1R57WmOk9aGPHxi7lpY78450HmPo7HfsfH9Lrp35GvDuklPytnxeF6_Jqnopi_kqaTlnMRGgLecgHHArkXHTOuushkxJzNGhtpnjSrvUKtQO8yzPhGy1ZDlwyYUVUwLn3zb4YQjomn3odiYcGmDNyUVzctGcXDQXF0fm4cx0iPi_P46VFpn4BT4KW3g |
| CODEN | ITIFA6 |
| Cites_doi | 10.1109/ICCV51070.2023.01909 10.1016/j.inffus.2020.06.014 10.1109/CVPR46437.2021.01268 10.1007/978-3-031-19781-9_9 10.1109/WIFS47025.2019.9035107 10.1109/TPAMI.2024.3350004 10.1609/aaai.v36i4.20391 10.1109/MIPR.2019.00103 10.1109/ICCV48922.2021.01383 10.1109/TPAMI.2022.3200384 10.1109/CVPR.2019.00241 10.1109/CVPR42600.2020.00813 10.48550/ARXIV.2106.12423 10.1109/CVPR42600.2020.01349 10.1109/TIT.1970.1054406 10.1016/j.patrec.2024.03.002 10.1109/IOLTS56730.2022.9897805 10.1109/ACCESS.2022.3151186 10.1109/ICASSP49357.2023.10094861 10.1109/CVPR42600.2020.00821 10.1109/CVPR.2019.00414 10.1109/ICCV48922.2021.01418 10.1109/CVPR.2019.00453 10.1109/ICCV48922.2021.00209 10.1007/978-3-030-58580-8_30 10.1109/CVPR52729.2023.01522 10.5244/C.31.42 10.1109/OJSP.2022.3190213 10.1109/CVPR52688.2022.01042 10.1109/ICCV.2015.425 10.1109/TPAMI.2012.256 10.1007/978-3-031-19806-9_21 10.1109/ICCV.2019.00765 10.1145/3544777 10.1609/aaai.v38i6.28385 10.1109/CVPR.2018.00366 10.1109/TPAMI.2021.3106743 10.1109/CVPRW59228.2023.00102 10.1109/CVPR42600.2020.00926 10.1504/IJMIC.2008.020996 10.1109/TNNLS.2022.3231924 10.48550/arXiv.1703.10717 |
| ContentType | Journal Article |
| DBID | 97E RIA RIE AAYXX CITATION |
| DOI | 10.1109/TIFS.2025.3592531 |
| DatabaseName | IEEE All-Society Periodicals Package (ASPP) 2005–Present IEEE All-Society Periodicals Package (ASPP) 1998–Present IEEE/IET Electronic Library CrossRef |
| DatabaseTitle | CrossRef |
| DatabaseTitleList | |
| 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 Computer Science |
| EISSN | 1556-6021 |
| EndPage | 8058 |
| ExternalDocumentID | 10_1109_TIFS_2025_3592531 11095738 |
| Genre | orig-research |
| GrantInformation_xml | – fundername: FOSTERER Project through Italian Ministry within the PRIN 2022 Program grantid: 202289RHHP – fundername: Security and Rights in the CyberSpace (SERICS) under the Ministero dell'Università e della Ricerca (MUR) National Recovery and Resilience Plan through European Union—NextGenerationEU grantid: PE00000014 – fundername: Remedy Project—New-Frontiers (PSR 2024) through MUR – fundername: China Scholarship Council (CSC) grantid: 202008370186 funderid: 10.13039/501100004543 |
| GroupedDBID | 0R~ 29I 4.4 5GY 5VS 6IK 97E AAJGR AASAJ AAWTH ABAZT ABQJQ ABVLG ACGFS ACIWK AENEX AETIX AGQYO AGSQL AHBIQ AKJIK AKQYR ALMA_UNASSIGNED_HOLDINGS ATWAV BEFXN BFFAM BGNUA BKEBE BPEOZ CS3 DU5 EBS EJD HZ~ IFIPE IPLJI JAVBF LAI M43 O9- OCL P2P PQQKQ RIA RIE RNS AAYXX CITATION |
| ID | FETCH-LOGICAL-c220t-317b2213f12b4e02acfbfb71854e9efe7b8f257f6b5e7fe989834c740912423b3 |
| IEDL.DBID | RIE |
| ISICitedReferencesCount | 0 |
| ISICitedReferencesURI | http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=001547648600016&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D |
| ISSN | 1556-6013 |
| IngestDate | Sat Nov 29 07:38:49 EST 2025 Sun Sep 28 03:48:02 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-c220t-317b2213f12b4e02acfbfb71854e9efe7b8f257f6b5e7fe989834c740912423b3 |
| ORCID | 0000-0002-0936-7904 0000-0002-7518-046X 0000-0002-7368-0866 |
| PageCount | 16 |
| ParticipantIDs | crossref_primary_10_1109_TIFS_2025_3592531 ieee_primary_11095738 |
| PublicationCentury | 2000 |
| PublicationDate | 20250000 2025-00-00 |
| PublicationDateYYYYMMDD | 2025-01-01 |
| PublicationDate_xml | – year: 2025 text: 20250000 |
| PublicationDecade | 2020 |
| PublicationTitle | IEEE transactions on information forensics and security |
| PublicationTitleAbbrev | TIFS |
| PublicationYear | 2025 |
| Publisher | IEEE |
| Publisher_xml | – name: IEEE |
| References | ref13 ref12 Yang (ref44) 2023 ref15 ref59 ref14 ref58 ref53 ref11 ref55 ref10 ref54 ref17 ref16 ref18 Yang (ref38) 2021 ref51 ref50 Gu (ref23) 2017 Liu (ref27) ref45 ref48 ref42 ref41 Zhang (ref46) Busbridge (ref57) Frank (ref39) Karras (ref56) ref8 Brock (ref47) ref9 ref3 ref6 ref5 ref40 ref35 ref34 Kim (ref36) ref31 ref30 ref33 Yu (ref37) 2020 ref32 Ho (ref7); 33 ref2 ref1 Karras (ref49) Fang (ref43) 2023 Vahdat (ref52) ref24 ref26 ref25 Vahdat (ref4) ref20 ref22 ref21 ref28 Vaze (ref19) ref60 Wang (ref29) |
| References_xml | – ident: ref41 doi: 10.1109/ICCV51070.2023.01909 – volume: 33 start-page: 6840 volume-title: Proc. NIPS ident: ref7 article-title: Denoising diffusion probabilistic models – start-page: 3247 volume-title: Proc. Int. Conf. Mach. Learn. ident: ref39 article-title: Leveraging frequency analysis for deep fake image recognition – start-page: 1 volume-title: Proc. Int. Conf. Learn. Represent. ident: ref46 article-title: Mixup: Beyond empirical risk minimization – ident: ref2 doi: 10.1016/j.inffus.2020.06.014 – start-page: 19667 volume-title: Proc. 34th Int. Conf. Neural Inf. Process. Syst. ident: ref4 article-title: NVAE: A deep hierarchical variational autoencoder – ident: ref53 doi: 10.1109/CVPR46437.2021.01268 – ident: ref15 doi: 10.1007/978-3-031-19781-9_9 – start-page: 1 volume-title: Proc. InInternational Conf. Learn. Represent. ident: ref36 article-title: Decentralized attribution of generative models – ident: ref8 doi: 10.1109/WIFS47025.2019.9035107 – ident: ref1 doi: 10.1109/TPAMI.2024.3350004 – ident: ref14 doi: 10.1609/aaai.v36i4.20391 – ident: ref11 doi: 10.1109/MIPR.2019.00103 – ident: ref40 doi: 10.1109/ICCV48922.2021.01383 – ident: ref17 doi: 10.1109/TPAMI.2022.3200384 – ident: ref33 doi: 10.1109/CVPR.2019.00241 – ident: ref5 doi: 10.1109/CVPR42600.2020.00813 – ident: ref6 doi: 10.48550/ARXIV.2106.12423 – ident: ref34 doi: 10.1109/CVPR42600.2020.01349 – ident: ref26 doi: 10.1109/TIT.1970.1054406 – ident: ref45 doi: 10.1016/j.patrec.2024.03.002 – ident: ref25 doi: 10.1109/IOLTS56730.2022.9897805 – ident: ref3 doi: 10.1109/ACCESS.2022.3151186 – ident: ref10 doi: 10.1109/ICASSP49357.2023.10094861 – ident: ref50 doi: 10.1109/CVPR42600.2020.00821 – ident: ref32 doi: 10.1109/CVPR.2019.00414 – year: 2017 ident: ref23 article-title: BadNets: Identifying vulnerabilities in the machine learning model supply chain publication-title: arXiv:1708.06733 – ident: ref13 doi: 10.1109/ICCV48922.2021.01418 – ident: ref54 doi: 10.1109/CVPR.2019.00453 – start-page: 1 volume-title: Proc. Adv. Neural Inf. Process. Syst. ident: ref57 article-title: How to scale your EMA – start-page: 1 volume-title: Proc. Int. Conf. Learn. Represent. (ICLR) ident: ref19 article-title: Open-set recognition: A good closed-set classifier is all you need – start-page: 12104 volume-title: Proc. 34th Int. Conf. Neural Inf. Process. Syst. ident: ref56 article-title: Training generative adversarial networks with limited data – ident: ref59 doi: 10.1109/ICCV48922.2021.00209 – ident: ref31 doi: 10.1007/978-3-030-58580-8_30 – ident: ref42 doi: 10.1109/CVPR52729.2023.01522 – start-page: 25033 volume-title: Proc. Adv. Neural Inf. Process. Syst. ident: ref29 article-title: OpenAUC: Towards AUC-oriented open-set recognition – ident: ref20 doi: 10.5244/C.31.42 – ident: ref22 doi: 10.1109/OJSP.2022.3190213 – year: 2021 ident: ref38 article-title: Learning to disentangle GAN fingerprint for fake image attribution publication-title: arXiv:2106.08749 – ident: ref51 doi: 10.1109/CVPR52688.2022.01042 – ident: ref55 doi: 10.1109/ICCV.2015.425 – ident: ref24 doi: 10.1109/TPAMI.2012.256 – start-page: 1 volume-title: Proc. Int. Conf. Learn. Represent. ident: ref47 article-title: Large scale GAN training for high fidelity natural image synthesis – ident: ref21 doi: 10.1007/978-3-031-19806-9_21 – start-page: 21464 volume-title: Proc. 34th Int. Conf. Neural Inf. Process. Syst. ident: ref27 article-title: Energy-based out-of-distribution detection – year: 2023 ident: ref44 article-title: Model synthesis for zero-shot model attribution publication-title: arXiv:2307.15977 – ident: ref12 doi: 10.1109/ICCV.2019.00765 – start-page: 11287 volume-title: Proc. NIPS ident: ref52 article-title: Score-based generative modeling in latent space – ident: ref60 doi: 10.1145/3544777 – ident: ref28 doi: 10.1609/aaai.v38i6.28385 – ident: ref30 doi: 10.1109/CVPR.2018.00366 – year: 2020 ident: ref37 article-title: Responsible disclosure of generative models using scalable fingerprinting publication-title: arXiv:2012.08726 – start-page: 1 volume-title: Proc. Int. Conf. Learn. Represent. ident: ref49 article-title: Progressive growing of GANs for improved quality, stability, and variation – ident: ref16 doi: 10.1109/TPAMI.2021.3106743 – ident: ref35 doi: 10.1109/CVPRW59228.2023.00102 – ident: ref58 doi: 10.1109/CVPR42600.2020.00926 – year: 2023 ident: ref43 article-title: Open set synthetic image source attribution publication-title: arXiv:2308.11557 – ident: ref9 doi: 10.1504/IJMIC.2008.020996 – ident: ref18 doi: 10.1109/TNNLS.2022.3231924 – ident: ref48 doi: 10.48550/arXiv.1703.10717 |
| SSID | ssj0044168 |
| Score | 2.4215152 |
| Snippet | With the continuous progress of AI technology, new generative architectures continuously appear, thus driving the attention of researchers towards the... |
| SourceID | crossref ieee |
| SourceType | Index Database Publisher |
| StartPage | 8043 |
| SubjectTerms | Analytical models Autoencoders backdoor injection Computer architecture deep learning for multimedia forensics Facial features Feature extraction Image forensics Image recognition open set recognition Prototypes Robustness Synthetic image attribution Training |
| Title | BOSC: A Backdoor-Based Framework for Open Set Synthetic Image Attribution |
| URI | https://ieeexplore.ieee.org/document/11095738 |
| Volume | 20 |
| WOSCitedRecordID | wos001547648600016&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: 1556-6021 dateEnd: 99991231 omitProxy: false ssIdentifier: ssj0044168 issn: 1556-6013 databaseCode: RIE dateStart: 20060101 isFulltext: true titleUrlDefault: https://ieeexplore.ieee.org/ providerName: IEEE |
| link | http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV3dS8MwED_c8EEfnM6J84s8-CR0a5OmWX3bxOJAprAJeytNegERW5md4H9v0g-dDz74VkJDw11zv7vk7n4Al5JLz9WJtoQmgeMbgHVCkShHUYraxdSlSVkofC9ms9FyGT7WxeplLQwilslnOLCP5V1-mqu1PSob2u6YXLBRC1pCBFWxVmN2DaxXdW-cB46JMlh9hWmmDBfTaG5CQcoHjIeUM-8XCG2wqpSgEnX-uZx92Ku9RzKu1H0AW5h1odMwM5B6o3Zhd6PN4CFMJw_zm2syJpNEvaR5vnImBrtSEjWJWcR4rsSmlpA5FmT-mRmv0HyBTF-NtSHj4psVqwdP0e3i5s6pKRSssN3CWFghKfWY9qj00QheaamlwSPuY4gahRxps2l1IDkKjZZMkvlKmKDPs46WZEfQzvIMj4EoLpDJgAoMA99LRej5IvVFYjxMKTRL-nDVyDR-qzplxGWE4YaxVUBsFRDXCuhDz8rz58ValCd_jJ_Cjp1enX2cQbtYrfEcttVH8fy-uih_hC_N_a8w |
| linkProvider | IEEE |
| linkToHtml | http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV3NS8MwFH_oFNSD0zlxfubgSehsk6ZZvW3DseKcQifsVpr2BUTcZHaC_71Jl-k8ePBWQtOW32veR_Le-wFcSi49V6XKEJoEjq8NrBOKNHMySlG5mLs0LQuFB2I4bI3H4aMtVi9rYRCxTD7Dprksz_LzaTY3W2XXpjsmF6y1DhuGOsuWay0Vrx5ZVL5xHjg6zmD2EFNPuh5FvVgHg5Q3GQ8pZ94vM7TCq1KalV71nx-0B7vWfyTthcD3YQ0nNaguuRmIXao12FlpNHgAUech7t6QNumk2Us-nc6cjrZeOektU7OI9l2JSS4hMRYk_pxov1C_gUSvWt-QdvHNi1WHp97tqNt3LImCgdsttI4VklKPKY9KHzX0mZJKaovEfQxRoZAtpZetCiRHodDQSTI_Ezrs84yrJdkhVCbTCR4BybhAJgMqMAx8Lxeh54vcF6n2MaVQLG3A1RLT5G3RKyMpYww3TIwAEiOAxAqgAXWD58-NFsrjP8YvYKs_uh8kg2h4dwLb5lGLnZBTqBSzOZ7BZvZRPL_Pzsuf4gshrbJ5 |
| 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=BOSC%3A+A+Backdoor-Based+Framework+for+Open+Set+Synthetic+Image+Attribution&rft.jtitle=IEEE+transactions+on+information+forensics+and+security&rft.au=Wang%2C+Jun&rft.au=Tondi%2C+Benedetta&rft.au=Barni%2C+Mauro&rft.date=2025&rft.issn=1556-6013&rft.eissn=1556-6021&rft.volume=20&rft.spage=8043&rft.epage=8058&rft_id=info:doi/10.1109%2FTIFS.2025.3592531&rft.externalDBID=n%2Fa&rft.externalDocID=10_1109_TIFS_2025_3592531 |
| thumbnail_l | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=1556-6013&client=summon |
| thumbnail_m | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=1556-6013&client=summon |
| thumbnail_s | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=1556-6013&client=summon |