DECENTRALISED REGISTRY FOR CONTENT ATTRIBUTION
THE PROJECT WILL DEVELOP A PROOF OF CONCEPT DECENTRALISED SERVICE FOR TRACING THE PROVENANCE OF MEDIA ASSETS (SUCH AS IMAGES, VIDEO, AUDIO)
The service will enable content publishers to lodge a copy of their asset (with associated provenance information) within an immutable decentralised storage. The service may then be queried by creators, publishers or content consumers who wish to determine the provenance, or origins, of a media asset. Applications will include solutions to fight misinformation (e.g. trace the provenance of images encountered online) and to determine the provenance of video (e.g. to decide whether to trust user generated content for inclusion in a new broadcast), or provenance of audio samples (e.g. in order to make decision around their use within a creative work). The project will leverage emerging standards for content attribution to embed provenance information within the asset metadata. The project will enable provenance about assets to be recalled using either cryptographic (exact) hashing or AI/content fingerprinting (soft hashing) technologies.
DLT technical challenge
Problem: How can we create a scalable decentralised infrastructure that enables storage and retrieval of asset provenance with high throughput, and without high transaction (‘gas’) fees. Are there appropriate technologies beyond DLT? What does the governance structure and trust model look like for these technologies?
Solution: Fingerprinting Develop tamper-evident content fingerprinting technologies for tracing images and video. Draw upon partner expertise for audio fingerprinting. Explore tamper visualization.
AI technical challenge
Problem: How can we compute robust content aware descriptors (hashes or fingerprints) to trace the provenance media assets such as images, audio and video or emerging forms of media such as 3D avatars, volumetric video or in-game digital assets such as fashion. How can we shape the user experience to communicate provenance of assets including where assets are tampered?
Solution: Platform Develop and deploy a DLT prototype across a federation of project partners, for storage and query of provenance information. Compare decentralised tech with a focus on Layer 2 DLT solutions and ZK rollups. Build a technical demonstrator of the platform e.g. for misinformation.
Problem: What are the privacy implications of a registry for asset attribution? What regulatory concerns might exist around such a service? What legal rights might provenance information imply?
Solution: Domain analysis of fake news Creating understanding of how civil society fights fake news and the emotional mechanism through which fake news spreads and can be counteracted.
Problem: What is the business model behind such a service – what is novel, and how is value created? What is the value proposition that incentivizes stakeholders to participate sustainably?
Solution: Regulatory space of provenance systems Report of the emerging legal landscape around AI and reverse asset lookup considering the privacy and IPR issues raised by a provenance system.