[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-83265-en":3,"doc-seo-83265-105":30,"detail-sidebar-cat-0-en-105":84},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},83265,13056703019662,"Evangeline","https://ap-avatar.wpscdn.com/avatar/be000253a8e92610077?_k=1778726343310543188",8,"Research & Report","Face-Trace: Open-Set Attribution and Progressive Discovery of Synthetic Face Generators","Advances in generative AI have produced increasingly realistic synthetic faces, creating new demands for multimedia forensics and reliable origin analysis. Traditional source attribution assumes a closed set of known generators, but real deployments face continuously emerging models and must organize rather than discard rejected samples. A pipeline is introduced combining known-generator classification with energy-based out-of-distribution rejection and unknown-generator discovery via clustering. Experiments on WILD report strong closed-set accuracy and meaningful open-set and incremental discovery results.","Face-trace: Open-Set Attribution and Progressive Discovery of  \nSynthetic Face Generators  \nAlessia Infantino, Claudio Schiavella Student Member IEEE, and Irene Amerini Member IEEE  \narXiv :2607 .07545v 1 [ cs .CV] 8 Jul 2026  \nAbstract—Recent advances in generative Artificial Intelligence have made synthetic face images increasingly realistic, creating new challenges for multimedia forensics. Source attribution methods should not only identify the generator of an image when the source is known, but also handle samples produced by previously unseen models. However, most existing approaches address synthetic face attribution in a closed-set setting, where all possible generators are available during training. This assumption does not hold in real-world scenarios, where new generators continuously appear and rejected samples should be organized rather than simply discarded.  \nIn this work we propose a pipeline for open-set synthetic face source attribution that combines known generator classification, energy-based OOD rejection, and unknown generator discovery. A classifier is trained on known generators using frozen I-JEPA embeddings, while rejected samples are represented by combining projected I-JEPA features with Forensic Self-Descriptors and then clustered to discover groups of unknown generators. We also extend the discovery stage to an incremental scenario, where rejected samples arrive over time.  \nExperiments on the WILD dataset show that the proposed method achieves 96.73% closed-set attribution accuracy. In the open-set setting, energy-based rejection reaches 71.25% balanced accuracy, while rejected samples are clustered into meaningful unknown-generator groups, obtaining an ARI of 0.81, an NMI of 0.90, and an overall clustering purity of 87.74% . In the incremental setting, the discovered generator space is progressively extended while maintaining a final purity of 99.23% . Cross-dataset experiments suggest that the pipeline can operate beyond the original dataset distribution, although post-processing remains challenging.  \nIndex Terms—Synthetic image attribution, Open-set recognition, Multimedia forensics, Novel category discovery, Deepfake.  \nI. INTRODUCTION  \nThe rapid advancement of Artificial Intelligence and the diffusion of generative models has made it possible to create synthetic images with a very high level of realism [1] . In particular, modern text-to-image models, GAN-based models, and diffusion models can generate human faces that look realistic and are difficult to distinguish from real ones [2] . While these technologies support legitimate applications in cinema, digital entertainment, advertising, and creative production, they also raise new challenges for multimedia forensics, where understanding the origin of a visual content is becoming increasingly important [3], [5] .  \nIn this context, it is important to distinguish between synthetic image detection and source attribution [4] . Synthetic  \nAll authors are with the Department of Computer, Control and Management Engineering (DIAG), Sapienza University of Rome, 00185 Rome, Italy (email: [infantino.1922069@studenti.uniroma1.it](infantino.1922069@studenti.uniroma1.it); [schiavella@diag.uniroma1.it](schiavella@diag.uniroma1.it);  \n[amerini@diag.uniroma1.it](amerini@diag.uniroma1.it)).  \nimage detection aims to understand whether an image is real or generated. Source attribution, instead, aims to identify which generative model produced the image. In this work, we focus on synthetic face source attribution, where the goal is to assign a synthetic face image to its generator.  \nMost existing attribution methods work in a closed-set scenario [32]. In this setting, the generators used during testing are the same generators seen during training. Therefore, each test image is assumed to belong to one of the known generator classes. New generative models are continuously released, and images may come from sources that were not available when the attributi","cbCaisXAak3eWIdR","https://ap.wps.com/l/cbCaisXAak3eWIdR","pdf",13832324,5,1,17,"English","en",105,"# Abstract\n# Introduction\n## Synthetic image detection vs. source attribution\n## Closed-set attribution and its limitations\n## Open-set attribution and rejection scores\n## Unknown-generator discovery and incremental discovery","[{\"question\":\"What evidence demonstrates effectiveness in open-set and incremental settings?\",\"answer\":\"On WILD, closed-set attribution reaches 96.73% accuracy, while open-set energy-based rejection achieves 71.25% balanced accuracy. Clustering of rejected samples yields an ARI of 0.81 and incremental discovery progressively extends the unknown-generator space while maintaining high final purity.\"}]",1784186379,43,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":79,"head_meta":81,"extra_data":83,"updated_unix":28},"face-trace-open-set-attribution-and-progressive-discovery-of-synthetic-face-generators","",{"@graph":36,"@context":78},[37,54,69],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/face-trace-open-set-attribution-and-progressive-discovery-of-synthetic-face-generators/83265/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":24,"description":14,"dateModified":62,"datePublished":63,"encodingFormat":61,"isAccessibleForFree":64,"interactionStatistic":65},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-07-25","2026-07-16",true,{"@type":66,"interactionType":67,"userInteractionCount":20},"InteractionCounter",{"@type":68},"ViewAction",{"@type":70,"mainEntity":71},"FAQPage",[72],{"name":73,"@type":74,"acceptedAnswer":75},"What evidence demonstrates effectiveness in open-set and incremental settings?","Question",{"text":76,"@type":77},"On WILD, closed-set attribution reaches 96.73% accuracy, while open-set energy-based rejection achieves 71.25% balanced accuracy. Clustering of rejected samples yields an ARI of 0.81 and incremental discovery progressively extends the unknown-generator space while maintaining high final purity.","Answer","https://schema.org",{"og:url":52,"og:type":80,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":82,"canonical":52},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":85},[86,90,94,98,102,107,112,115,120,123,127],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":87,"show_sort_weight":88,"slug":89},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":91,"show_sort_weight":92,"slug":93},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":95,"show_sort_weight":96,"slug":97},"Exam",70,"exam",{"id":20,"doc_module":4,"doc_module_name":46,"category_name":99,"show_sort_weight":100,"slug":101},"Comic",60,"comic",{"id":103,"doc_module":4,"doc_module_name":46,"category_name":104,"show_sort_weight":105,"slug":106},6,"Technology",50,"technology",{"id":108,"doc_module":4,"doc_module_name":46,"category_name":109,"show_sort_weight":110,"slug":111},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":113,"slug":114},30,"research-report",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},9,"Religion & Spirituality",20,"religion-spirituality",{"id":118,"doc_module":4,"doc_module_name":46,"category_name":121,"show_sort_weight":118,"slug":122},"World Cup","world-cup",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":124,"slug":126},10,"Lifestyle","lifestyle",{"id":128,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":20,"slug":130},19,"General","general"]