Ekila: synthetic media provenance and attribution for generative art

Kar Balan[1], Shruti Agarwal[2], Simon Jenni[2], Andy Parsons [2], Andrew Gilbert[1], John Collomosse[1,2]
Unviersity of Surrey [1], Adobe Research [2]
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2023
Paper Supplementary Poster

EKILA combines robust visual attribution with Distributed Ledger Technology (DLT) to recognize and reward creative contributions to generative art. A cymbal generated by a Latent Diffusion Model (LDM) trained on LAION-400M is attributed to a subset of training images, credit weight apportioned, and royalties paid using our proposed method.

Abstract

We present EKILA; a decentralized framework that enables creatives to receive recognition and reward for their contributions to generative AI (GenAI). EKILA proposes a robust visual attribution technique and combines this with an emerging content provenance standard (C2PA) to address the problem of synthetic image provenance -- determining the generative model and training data responsible for an AI-generated image. Furthermore, EKILA extends the non-fungible token (NFT) ecosystem to introduce a tokenized representation for rights, enabling a triangular relationship between the asset's Ownership, Rights, and Attribution (ORA). Leveraging the ORA relationship enables creators to express agency over training consent and, through our attribution model, to receive apportioned credit, including royalty payments for the use of their assets in GenAI.

Ekila is a flexible model, integrating C2PA manifests with NFT to trace dynamic ownership of training data assets. Training images are minted as NFTs and carry a link (ARA) to each NFT within the C2PA metadata of the asset. The C2PA metadata thus provides a way to determine the current owner of an asset, including their wallet address, for receipt of royalties. This obviates the need to store the wallet address statically in the ingredient manifests.

Poster

BibTeX

@inproceedings{Balan:EKila:CVPRWS:2023,
        AUTHOR = Balan, Kar and Agarwal, Shruti and Jenni, Simon  and Parsons, Andy and  Gilbert Andrew and Collomosse, John",
        TITLE = "Ekila: synthetic media provenance and attribution for generative art",
        BOOKTITLE = "IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops",
        YEAR = "2023",
        ​}