CGI-DM: Digital Copyright Authentication for Diffusion Models via Contrasting Gradient Inversion
Diffusion Models (DMs) have evolved into advanced image generation tools, especially for few-shot generation where a pretrained model is fine-tuned on a small set of images to capture a specific style or object. Despite their success, concerns exist about potential copyright violations stemming from...
Gespeichert in:
Veröffentlicht in: | arXiv.org 2024-03 |
---|---|
Hauptverfasser: | , , , , , , |
Format: | Artikel |
Sprache: | eng |
Schlagworte: | |
Online-Zugang: | Volltext |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
container_end_page | |
---|---|
container_issue | |
container_start_page | |
container_title | arXiv.org |
container_volume | |
creator | Wu, Xiaoyu Yang, Hua Liang, Chumeng Zhang, Jiaru Wang, Hao Song, Tao Guan, Haibing |
description | Diffusion Models (DMs) have evolved into advanced image generation tools, especially for few-shot generation where a pretrained model is fine-tuned on a small set of images to capture a specific style or object. Despite their success, concerns exist about potential copyright violations stemming from the use of unauthorized data in this process. In response, we present Contrasting Gradient Inversion for Diffusion Models (CGI-DM), a novel method featuring vivid visual representations for digital copyright authentication. Our approach involves removing partial information of an image and recovering missing details by exploiting conceptual differences between the pretrained and fine-tuned models. We formulate the differences as KL divergence between latent variables of the two models when given the same input image, which can be maximized through Monte Carlo sampling and Projected Gradient Descent (PGD). The similarity between original and recovered images serves as a strong indicator of potential infringements. Extensive experiments on the WikiArt and Dreambooth datasets demonstrate the high accuracy of CGI-DM in digital copyright authentication, surpassing alternative validation techniques. Code implementation is available at https://github.com/Nicholas0228/Revelio. |
format | Article |
fullrecord | <record><control><sourceid>proquest</sourceid><recordid>TN_cdi_proquest_journals_2968619786</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2968619786</sourcerecordid><originalsourceid>FETCH-proquest_journals_29686197863</originalsourceid><addsrcrecordid>eNqNi90KgjAARkcQJOU7DLoWdObU7kLLvPCuexu56UQ224_Q2zehB-jq4_CdswEeiuMoyE4I7YCv9RiGIcIpSpLYA8-iqoOyOcOS99yQCRZy_ijeDwZerBmoMPxFDJcCMqmcxJjVKzWyo5OGCyeuEEYRbbjoYaVIx10Ea7FQtZoHsGVk0tT_7R4cb9dHcQ9mJd-WatOO0irhrhblOMNRnmY4_s_6AocORKw</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2968619786</pqid></control><display><type>article</type><title>CGI-DM: Digital Copyright Authentication for Diffusion Models via Contrasting Gradient Inversion</title><source>Free E- Journals</source><creator>Wu, Xiaoyu ; Yang, Hua ; Liang, Chumeng ; Zhang, Jiaru ; Wang, Hao ; Song, Tao ; Guan, Haibing</creator><creatorcontrib>Wu, Xiaoyu ; Yang, Hua ; Liang, Chumeng ; Zhang, Jiaru ; Wang, Hao ; Song, Tao ; Guan, Haibing</creatorcontrib><description>Diffusion Models (DMs) have evolved into advanced image generation tools, especially for few-shot generation where a pretrained model is fine-tuned on a small set of images to capture a specific style or object. Despite their success, concerns exist about potential copyright violations stemming from the use of unauthorized data in this process. In response, we present Contrasting Gradient Inversion for Diffusion Models (CGI-DM), a novel method featuring vivid visual representations for digital copyright authentication. Our approach involves removing partial information of an image and recovering missing details by exploiting conceptual differences between the pretrained and fine-tuned models. We formulate the differences as KL divergence between latent variables of the two models when given the same input image, which can be maximized through Monte Carlo sampling and Projected Gradient Descent (PGD). The similarity between original and recovered images serves as a strong indicator of potential infringements. Extensive experiments on the WikiArt and Dreambooth datasets demonstrate the high accuracy of CGI-DM in digital copyright authentication, surpassing alternative validation techniques. Code implementation is available at https://github.com/Nicholas0228/Revelio.</description><identifier>EISSN: 2331-8422</identifier><language>eng</language><publisher>Ithaca: Cornell University Library, arXiv.org</publisher><subject>Copyright ; Digital imaging ; Image processing</subject><ispartof>arXiv.org, 2024-03</ispartof><rights>2024. This work is published under http://creativecommons.org/licenses/by/4.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>780,784</link.rule.ids></links><search><creatorcontrib>Wu, Xiaoyu</creatorcontrib><creatorcontrib>Yang, Hua</creatorcontrib><creatorcontrib>Liang, Chumeng</creatorcontrib><creatorcontrib>Zhang, Jiaru</creatorcontrib><creatorcontrib>Wang, Hao</creatorcontrib><creatorcontrib>Song, Tao</creatorcontrib><creatorcontrib>Guan, Haibing</creatorcontrib><title>CGI-DM: Digital Copyright Authentication for Diffusion Models via Contrasting Gradient Inversion</title><title>arXiv.org</title><description>Diffusion Models (DMs) have evolved into advanced image generation tools, especially for few-shot generation where a pretrained model is fine-tuned on a small set of images to capture a specific style or object. Despite their success, concerns exist about potential copyright violations stemming from the use of unauthorized data in this process. In response, we present Contrasting Gradient Inversion for Diffusion Models (CGI-DM), a novel method featuring vivid visual representations for digital copyright authentication. Our approach involves removing partial information of an image and recovering missing details by exploiting conceptual differences between the pretrained and fine-tuned models. We formulate the differences as KL divergence between latent variables of the two models when given the same input image, which can be maximized through Monte Carlo sampling and Projected Gradient Descent (PGD). The similarity between original and recovered images serves as a strong indicator of potential infringements. Extensive experiments on the WikiArt and Dreambooth datasets demonstrate the high accuracy of CGI-DM in digital copyright authentication, surpassing alternative validation techniques. Code implementation is available at https://github.com/Nicholas0228/Revelio.</description><subject>Copyright</subject><subject>Digital imaging</subject><subject>Image processing</subject><issn>2331-8422</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2024</creationdate><recordtype>article</recordtype><sourceid>ABUWG</sourceid><sourceid>AFKRA</sourceid><sourceid>AZQEC</sourceid><sourceid>BENPR</sourceid><sourceid>CCPQU</sourceid><sourceid>DWQXO</sourceid><recordid>eNqNi90KgjAARkcQJOU7DLoWdObU7kLLvPCuexu56UQ224_Q2zehB-jq4_CdswEeiuMoyE4I7YCv9RiGIcIpSpLYA8-iqoOyOcOS99yQCRZy_ijeDwZerBmoMPxFDJcCMqmcxJjVKzWyo5OGCyeuEEYRbbjoYaVIx10Ea7FQtZoHsGVk0tT_7R4cb9dHcQ9mJd-WatOO0irhrhblOMNRnmY4_s_6AocORKw</recordid><startdate>20240317</startdate><enddate>20240317</enddate><creator>Wu, Xiaoyu</creator><creator>Yang, Hua</creator><creator>Liang, Chumeng</creator><creator>Zhang, Jiaru</creator><creator>Wang, Hao</creator><creator>Song, Tao</creator><creator>Guan, Haibing</creator><general>Cornell University Library, arXiv.org</general><scope>8FE</scope><scope>8FG</scope><scope>ABJCF</scope><scope>ABUWG</scope><scope>AFKRA</scope><scope>AZQEC</scope><scope>BENPR</scope><scope>BGLVJ</scope><scope>CCPQU</scope><scope>DWQXO</scope><scope>HCIFZ</scope><scope>L6V</scope><scope>M7S</scope><scope>PIMPY</scope><scope>PQEST</scope><scope>PQQKQ</scope><scope>PQUKI</scope><scope>PRINS</scope><scope>PTHSS</scope></search><sort><creationdate>20240317</creationdate><title>CGI-DM: Digital Copyright Authentication for Diffusion Models via Contrasting Gradient Inversion</title><author>Wu, Xiaoyu ; Yang, Hua ; Liang, Chumeng ; Zhang, Jiaru ; Wang, Hao ; Song, Tao ; Guan, Haibing</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-proquest_journals_29686197863</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2024</creationdate><topic>Copyright</topic><topic>Digital imaging</topic><topic>Image processing</topic><toplevel>online_resources</toplevel><creatorcontrib>Wu, Xiaoyu</creatorcontrib><creatorcontrib>Yang, Hua</creatorcontrib><creatorcontrib>Liang, Chumeng</creatorcontrib><creatorcontrib>Zhang, Jiaru</creatorcontrib><creatorcontrib>Wang, Hao</creatorcontrib><creatorcontrib>Song, Tao</creatorcontrib><creatorcontrib>Guan, Haibing</creatorcontrib><collection>ProQuest SciTech Collection</collection><collection>ProQuest Technology Collection</collection><collection>Materials Science & Engineering Collection</collection><collection>ProQuest Central (Alumni Edition)</collection><collection>ProQuest Central UK/Ireland</collection><collection>ProQuest Central Essentials</collection><collection>ProQuest Central</collection><collection>Technology Collection</collection><collection>ProQuest One Community College</collection><collection>ProQuest Central Korea</collection><collection>SciTech Premium Collection</collection><collection>ProQuest Engineering Collection</collection><collection>Engineering Database</collection><collection>Publicly Available Content Database</collection><collection>ProQuest One Academic Eastern Edition (DO NOT USE)</collection><collection>ProQuest One Academic</collection><collection>ProQuest One Academic UKI Edition</collection><collection>ProQuest Central China</collection><collection>Engineering Collection</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Wu, Xiaoyu</au><au>Yang, Hua</au><au>Liang, Chumeng</au><au>Zhang, Jiaru</au><au>Wang, Hao</au><au>Song, Tao</au><au>Guan, Haibing</au><format>book</format><genre>document</genre><ristype>GEN</ristype><atitle>CGI-DM: Digital Copyright Authentication for Diffusion Models via Contrasting Gradient Inversion</atitle><jtitle>arXiv.org</jtitle><date>2024-03-17</date><risdate>2024</risdate><eissn>2331-8422</eissn><abstract>Diffusion Models (DMs) have evolved into advanced image generation tools, especially for few-shot generation where a pretrained model is fine-tuned on a small set of images to capture a specific style or object. Despite their success, concerns exist about potential copyright violations stemming from the use of unauthorized data in this process. In response, we present Contrasting Gradient Inversion for Diffusion Models (CGI-DM), a novel method featuring vivid visual representations for digital copyright authentication. Our approach involves removing partial information of an image and recovering missing details by exploiting conceptual differences between the pretrained and fine-tuned models. We formulate the differences as KL divergence between latent variables of the two models when given the same input image, which can be maximized through Monte Carlo sampling and Projected Gradient Descent (PGD). The similarity between original and recovered images serves as a strong indicator of potential infringements. Extensive experiments on the WikiArt and Dreambooth datasets demonstrate the high accuracy of CGI-DM in digital copyright authentication, surpassing alternative validation techniques. Code implementation is available at https://github.com/Nicholas0228/Revelio.</abstract><cop>Ithaca</cop><pub>Cornell University Library, arXiv.org</pub><oa>free_for_read</oa></addata></record> |
fulltext | fulltext |
identifier | EISSN: 2331-8422 |
ispartof | arXiv.org, 2024-03 |
issn | 2331-8422 |
language | eng |
recordid | cdi_proquest_journals_2968619786 |
source | Free E- Journals |
subjects | Copyright Digital imaging Image processing |
title | CGI-DM: Digital Copyright Authentication for Diffusion Models via Contrasting Gradient Inversion |
url | https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-23T02%3A28%3A23IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest&rft_val_fmt=info:ofi/fmt:kev:mtx:book&rft.genre=document&rft.atitle=CGI-DM:%20Digital%20Copyright%20Authentication%20for%20Diffusion%20Models%20via%20Contrasting%20Gradient%20Inversion&rft.jtitle=arXiv.org&rft.au=Wu,%20Xiaoyu&rft.date=2024-03-17&rft.eissn=2331-8422&rft_id=info:doi/&rft_dat=%3Cproquest%3E2968619786%3C/proquest%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_pqid=2968619786&rft_id=info:pmid/&rfr_iscdi=true |