I've thought of something to do this automagically during previous protests. in the event that a picture is leaked that contains someone's face in an identifiable manner, it'll do a basic facial detection run on it, and then use a GAN to generate new faces to plaster over those who might be identified. It'll them spam upload as many different versions of the pic literally anywhere it can. Goal being to muddy the waters. The idea not being to try to convince people its someone else, but rather to give plausible deniability that the suspect's face isn't the actual original one and is one of the photoshopped varients.

Thoughts? It'd be super easy to do, but I feel the premise is dubious at best and at worst could draw additional unwanted attention to the original fotos