The objective was to create a mobile application that will allow the authentication of photos taken through the application, verify the authenticity of the photos that already exist and show their history, so that it is evident in what way they have been manipulated from the moment they were taken.
Solution
In addition to the Android application, an SDK (Software Development Kit) was also created in Kotlin and Rust languages. This SDK provides a “plug and play” interface for potential use by other applications and acts as a superstructure on top of CAI's C2PA implementing library, an open-source format that guarantees media authenticity, supported by many well-known companies (Adobe, Microsoft, BBC, Sony,...). When we created the application, the C2PA library began to be created in the form of “proof of concept” and we worked directly with its creators.
On the application side, we tested the possibilities of obtaining relevant information from the mobile device that would help us identify the photo and allow us to assess its authenticity. As part of this project, we also analyzed what data is deleted from popular social networks and communication applications. On the server side, we analyzed and implemented various metrics and heuristics (such as metadata consistency, device integrity, estimated location) with which we subsequently evaluated the authenticity of the photos.
Challenges and learnings
Since it was R&D, we became researchers. We took the investor's idea: “I would like there to be no fraudulent photos on the Internet, or at least to distinguish them from real ones.” And from scratch we designed a system that had the potential to change that.