C2PA-Related Papers
Introduce C2PA
- This Election Year, Look for Content Credentials: Media organizations combat deepfakes and disinformation with digital manifests, Eliza Strickland, IEE Spectrum, 2024
- talking about what C2PA is and how the adoption may work
- Understanding the Impact of AI-Generated Deepfakes on Public Opinion, Political Discourse, and Personal Security in Social Media, Prakash L. Kharvi, IEEE SP, 2024
- why policies & companies should adopt C2PA
- Defining best practices for opting out of ML training, Paul Keller, Zuzanna Warso, Open Future policy brief, 2023
- C2PA has entry to opt out of ML training
- AI-Generated Images as an Emergent Record Format, Jessica Bushey, IEEE BigData, 2023
- literature review by computational archival science (CAS) people, ITrustAI
- GenAI literature in medicine & law enforcement & journalism
- copyright attribution; public trust in democracy
- Confidence-Building Measures for Artificial Intelligence: Workshop Proceedings, Sarah Shoker, Andrew Reddie, et al., arXiv, 2023
- foundational models can be used for insecure things
- watermarking has tempering and adoption problems
- ⭐ Solutions to Deepfakes: Can Camera Hardware, Cryptography, and Deep Learning Verify Real Images?, Alexander Vilesov, Yuan Tian, Nader Sehatbakhsh, Achuta Kadambi, arXiv, 2024
- ways to verify images, including C2PA
- need every camera to support C2PA
- can spoof by taking photo of a photo
- prevent by various ways to distinguish 2D from 3D: camera structure from motion, hand jitter, stereo-based depth triangulation, lidar sensing, polarization-based sensing, time-of-flight sensing, spatially altered Bayer pattern, sparse polarization-sensitive pixel sampling
- useless if object is already flat
- new file format for “real” images
- Global content revocation on the internet: a case study in technology ecosystem transformation, Narek Galstyan, James McCauley, Hany Farid, Sylvia Ratnasamy, Scott Shenker, HotNets, 2022
- wild new idea to rethink content revocation
- Deepfake Fraud Detection: Safeguarding Trust in Generative Ai, Felipe Romero-Moreno, SSRN Preprint, 2024
- from law person perspective, C2PA and other tools to combat deepfake
Enhance or extend C2PA
- Ensuring privacy in provenance information for images, Nikolaos Fotos, Jaime Delgado, IEEE DSP, 2023; Towards Privacy-Enhancing Provenance Annotations for Images, Nikolaos Fotos, Jaime Delgado, IEEE ICIP, 2024
- add privacy-preserving features
- A Blockchain based Framework for Content Provenance and Authenticity, Emil Bureacă, Iulian Aciobăniței, IEEE ECAI, 2024
- blockchain on top of C2PA
- Integrating Content Authenticity with DASH Video Streaming, Stefano Petrangeli, Haoliang Wang, Maurice Fisher, Dave Kozma, Massy Mahamli, Pia Blumenthal, Andy Parsons, ACM MMSys, 2024
- C2PA in video stream
- Trust Nobody: Privacy-Preserving Proofs for Edited Photos with Your Laptop, Pierpaolo Della Monica, Ivan Visconti, Andrea Vitaletti, Marco Zecchini, IEEE SP, 2025; Trust Nobody: Privacy-Preserving Proofs for Edited Photos with Your Laptop, Pierpaolo Della Monica, Ivan Visconti, Andrea Vitaletti, Marco Zecchini, Cryptology ePrint Archive, 2024
- privacy-preserving proofs that an image is manipulation of another
- VerITAS: Verifying Image Transformations at Scale, Trisha Datta, Binyi Chen, Dan Boneh, Cryptology ePrint Archive, 2024; VIMz: Verifiable Image Manipulation using Folding-based zkSNARKs, Stefan Dziembowski, Shahriar Ebrahimi, Parisa Hassanizadeh, Cryptology ePrint Archive, 2024
- zero-knowledge proof that an image is transformation of another
- To Authenticity, and Beyond! Building Safe and Fair Generative AI Upon the Three Pillars of Provenance, John Collomosse, Andy Parsons, IEEE CGA, 2024
- three pillars: metadata, fingerprint, watermark
- Pseudorandom Error-Correcting Codes, Miranda Christ, Sam Gunn, IACR, 2024
- robust codeword appear random w/o decoding key
- -hardness from learning parity with noise (LPN) problem (NP-hard)
- undetectable watermark & tolerate alternation
- steganography
- robust codeword appear random w/o decoding key
Applying C2PA
- EKILA: Synthetic Media Provenance and Attribution for Generative Art, Kar Balan, Shruti Agarwal, Simon Jenni, Andy Parsons, Andrew Gilbert, John Collomosse, IEEE CVPR, 2023; DECORAIT - DECentralized Opt-in/out Registry for AI Training, Kar Balan, Andrew Gilbert, Alexander Black, Simon Jenni, Andy Parsons, John Collomosse, ACM CVMP, 2023
- blockchain to index & attribute GenAI contribution by fingerprinting
- C2PA to express consent & payment preference
- Interoperable Provenance Authentication of Broadcast Media using Open Standards-based Metadata, Watermarking and Cryptography, John C. Simmons, Joseph M. Winograd, IBC 2024
- C2PA & Advanced Television Systems Committee (ATSC) for broadcast provenance
- ATSC: transfer metadata over broadcast
Alternatives or complements to C2PA
- Towards Trustworthy Digital Media In The Aigc Era: An Introduction To The Upcoming IsoJpegTrust Standard, Jiayun Mo, Xin Kang, Ziyuan Hu, Haibo Zhou, Tieyan Li, Xiaojun Gu, IEEE COMSTD, 2023; An International Standard For Assessing Trustworthiness In Media, Deepayan Bhowmik, Sabrina Caldwell, Jaime Delgado, Touradj Ebrahimi, Nikolaos Fotos, Xiaojun Gu, IEEE ICIP, 2024
- ISO JPEG Trust: “trust profile” vs “trust credential” question-answer validation
- emphasis on trustworthiness
- TRAIT: A Trusted Media Distribution Framework, James Rainey, Mohamed Elawady, Charith Abhayartne, Deepayan Bhowmik, IEEE DSP, 2023
- blockchain to detect media manipulation
Scaling PKI
- A Blockchain-Based PKI Management Framework, Alexander Yakubov, Wazen M. Shbair, Anders Wallbom, David Sanda, Radu State, IEEE/IFIP Man2Block, 2018
- PKI insecure when CA compromised
- blockchain-based PKI
Fake image detection
- Can people identify original and manipulated photos of real-world scenes?, Sophie J Nightingale, Kimberley A Wade, Derrick G Watson, Springer Cognitive Research, 2017
- people can hardly tell if and how image manipulated
- Explaining Why Fake Photos are Fake: Does It Work?, Margie Ruffin, Gang Wang, Kirill Levchenko, ACM HCI, 2022
- people cannot tell if image manipulated
- explaining manipulation does not always help
- AI-synthesized faces are indistinguishable from real faces and more trustworthy, Sophie J. Nightingale, Hany Farid, PNAS, 2022
- people cannot tell if face is AI-generated
- Seeing is not always believing: Benchmarking Human and Model Perception of AI-Generated Images, Zeyu Lu, Di Huang, Lei Bai, Jingjing Qu, Chengyue Wu, Xihui Liu, Wanli Ouyang, NeurIPS, 2023
- accuracy detecting fake photorealistic image: human 60%, best open model 87%
- but model high false positive; best overall accuracy 83%
- 2 million fake image for model benchmark
- accuracy detecting fake photorealistic image: human 60%, best open model 87%
- Organic or Diffused: Can We Distinguish Human Art from AI-generated Images?, Anna Yoo Jeong Ha, Josephine Passananti, Ronik Bhaskar, Shawn Shan, Reid Southen, Haitao Zheng, Ben Y. Zhao, ACM CCS, 2024
- ML detector & human cannot tell if image is AI-generated
- ways to trick ML detector
- ML detector trained on specific generator not good at new generator
- human expert better than ordinary human
- Fake It Until You Break It: On the Adversarial Robustness of AI-generated Image Detectors, Sina Mavali, Jonas Ricker, David Pape, Yash Sharma, Asja Fischer, Lea Schönherr, arXiv, 2024
- perturbation can make forensic classifier useless
- best model DRCT-CLIP has 88% accuracy
- DRCT: Diffusion Reconstruction Contrastive Training towards Universal Detection of Diffusion Generated Images, Baoying Chen, Jishen Zeng, Jianquan Yang, Rui Yang, ICML, 2024
Anti-training
- GLAZE: Protecting Artists from Style Mimicry by Text-to-Image Models, Shawn Shan, Jenna Cryan, Emily Wenger, Haitao Zheng, Rana Hanocka, Ben Y. Zhao USENIX Security, 2023
- add noise to image to brick ML model from generating similar image
Adoption of C2PA-like standards
- Is The Web HTTP/2 Yet?, Matteo Varvello, Kyle Schomp, David Naylor, Jeremy Blackburn, Alessandro Finamore, Konstantina Papagiannaki, PAM, 2016
- crawl & track HTTP 2 usage
Last edited: