This is How Adobe’s Upcoming Photo ‘Authenticity’ System Will Work

Almost 9 months after announcing the so-called Content Authenticity Initiative (CAI) for preventing image theft and manipulation online, Adobe has finally released details on how this special authentication system will work when they begin rolling it out later this year.

First announced at AdobeMAX 2019 last November, the CAI is a system for permanently attaching attribution and other metadata to an image in order to combat misinformation (i.e. sneaky photo manipulation) and photo theft. However, when it was first announced, details about how it would actually work were pretty sparse.

In a new white paper about the CAI, Adobe finally dived into the details, describing a type of metadata that is “cryptographically sealed and verifiable” so that you can see exactly who captured an image and how it has been altered all the way from “creation to consumption.” This CAI attribution data will consist of “assertions” and “claims.” Assertions contain information about image creation, authorship, and edit actions, which are ultimately wrapped up into “verifiable units” called claims.

Here’s a brief explanation from the white paper itself:

Assertions are cryptographically hashed and their hashes are gathered together into a claim. A claim is a digitally signed data structure that represents a set of assertions along with one or more cryptographic hashes on the data of an asset. The signature ensures the integrity of the claim and makes the system tamper-evident[…]

Each time the asset reaches a specific key point in its lifecycle, such as initial creation, completion of some editing operations, publication to social media, etc. a new set of assertions and a claim are created. Each new claim refers to the previous claim, thus creating a chain of provenance for the asset

CAI specifications will be available under an open source license, but Adobe says they’re having “really active conversations across the board with lots of different companies” about implementing the system.

And that, ultimately, will be the challenge here. As The Verge points out, the success of CAI as a standard will depend on hardware and software companies cooperating: camera and smartphone makers would need to implement a system for embedding CAI data into photos as they’re captured and edited, software companies (including Adobe’s competitors) would need to implement a new CAI tag every time an edit is made, and publishers/social media sites would need to display CAI information prominently enough that people actually take note, without overwhelming users.

The hope is that the system becomes so wide-spread that any photo without CAI data attached will be viewed with increased skepticism, while CAI images will be seen as above-board and otherwise “manipulation proof.” That’s a tall order…

As Allen Murabayashi wrote in November when the standard was first announced, there are reasons to be skeptical that this will have a significant impact in real life. Metadata standards for attaching copyright holder information already exist, and Adobe would need widespread buy-in for their system to work on an Internet-wide scale.

However, the details released today should give photographers hope that Adobe will pull this off. If they can convince major hardware and software manufacturers to play ball, and loop in the social media giants while they’re at it, this could significantly tamp down on wanton image theft and manipulation on the World Wide Web. Wouldn’t that be nice…

To dive into all of the details and read the full white paper, click here. No official launch date is set as of yet, but Adobe plans to release a prototype CAI tagging system to a “subset” of Photoshop and Behance users by the end of 2020.