Dall-E 3 Generated Image Recognition

In the following article I read something about that Dall-E 3 Team has an internal tool that is being developed for recognizing images that are made/generated by Dall-E 3.

Title: OpenAI kondigt Dall-E 3 aan, wordt geïntegreerd in ChatGPT
Platform: Tweakers

I wanted to contribute to this solution by sharing my idea.

Like a QR code that is unique, you could use very small pixels (should not be visible to human eyes) within the image in such a way that these pixels are unique and distinguishable of each other.

This might be an identification number that is known to the Dall-E 3 Team. Thus, it should generated in such a way that after validating the unique number with custom algorithms they should be recognized as ‘Image made with Dall-E 3’. Having these custom algorithms only known to the Dall-E 3 team would mean that these unique identifiers cannot be generated by others to sabotage generated images.

Location for these unique identifiers would be te center of an image, since sometimes edges of images are being cut off and the center always remains robust.

By that the team just could search for these pixels/unique identifiers within the picture. I think that it would be very easy through this approach to identify whether or not an image has been generated by Dall-E 3.

Hopefully others can also share their ideas and opinions about my idea.

Have a great day,