Content Authenticity Initiative combats photographic mis/disinformation by proving ‘what is real,’ ISOJ workshop hears


Family members love to share pictures and funny posts that they find interesting on Facebook. But when images seem outrageously untrue, like pyramids in Antarctica, how can we tell where the photo really came from?

The new Adobe-led Content Authenticity Initiative is working to fight this form of mis/disinformation with a new tool: provenance. According to the Head of the Content Authenticity Initiative, Santiago Lyon, the project seeks to fight disinformation by providing “content credentials,” a way to show a file’s changes and origins, the result of a collaboration of multiple corporations focused on media transparency.

After first trying forms of detection, policy change and media literacy education, the idea of provenance was adopted as the main goal. In a world of fast-changing technology, provenance works to “prove what’s real as opposed to detecting what’s false,” Lyon, an award-winning photojournalist, told a breakfast workshop April 15 at the 24th ISOJ.

“We’re seeking to modify the way people interact with, or provide tools that allow people to modify the way they interact with content,” Lyon said. “And we’re also seeking to sort of get back to this notion, I suppose, of checking your source and checking, you know, asking questions, being skeptical. Who’s behind that? What media organization is that?“

The Content Authenticity Initiative (CAI) establishes provenance in sequential steps:

First, capture establishes secure metadata for any new photos. This metadata contains information about where and when the photo was taken, but the creator can choose to remain anonymous for any reason. It is also cryptographed in a way that any tampering attempts would be shown.

The option of anonymity could prove important for sources and reporters who would be endangered by revealing their location, something journalists in the audience seemed especially interested in.

“A journalist working in a war zone would be ill advised to share their GPS location, because as we’ve seen in the past, that can prove fatal, or a human rights defender working in a repressive regime” Lyon said. “The last thing they want probably is to have their identity … associated with a particular piece of damning evidence of human rights abuses. So all of this is customizable according to the needs of the user.”

Second, the editing stage. CAI is working to integrate technology into content editing systems, like Adobe, so that when a file is edited new metadata will be created showing the edit history.

Finally, in the publishing stage, news organizations and content creators can choose to display the information regarding the origin and history of the image.

While the first step is capture, the hardware to do this has not been released, though there are models currently being tested. The CAI model of provenance, however, can begin at either the capture or edit stages.

“But if we wait for the camera manufacturers to create production devices, and we wait for the software to become more ubiquitous, we’ll be sitting on our hands for too long, and we think it’s urgent. So this is a way of adding simple content credentials to all of the outgoing content,” Lyon said.

The project currently only authenticates photos and does not have a uniform display system, although they are in partnership with a few news outlets to test display options.

“We’re at the beginning of this journey, but we’re looking to have real time implementation sometime towards the end of this year, and get this in front of the general public, and then build on that,” Lyon said. Eventually, they will tackle the challenge of video authentication.

In addition to creating the Content Authenticity Initiative, Adobe worked with Microsoft to create the Coalition for Content Provenance and Authenticity, or C2PA, which is working to create international standards for using provenance — kind of like a best practice for all the different companies that might use or create provenance systems.

“One way to think about the C2PA is, they are the architects,” Lyon said. “So the C2PA is really developing the technical standard, or the blueprint on which everything is based. And the Content Authenticity Initiative is the construction company, taking that blueprint and making open source tools with it that everybody can use.”

Currently, the Initiative offers three options for integrating their provenance model into content creation and distribution: a Java Script SDK, which has only data reading capabilities, a C2PA command line tool, which can read and write small amounts of the metadata, and the Rust SDK, a full SDK allowing users to read full metadata files and inspect them.

Content viewers can also inspect the CAI data available for any file using the Verify feature to upload and inspect an image.

Author’s Bio: Molly-Jo Tilton is a third-year journalism student. She is the current Audio Editor for The Daily Texan and a former editorial intern at Austin Woman Magazine.