How to bypass face recognition in a photo?

Question

Neural networks

Are there any ways to process photos in such a way that the neural network, trained to look for matches by faces, could not find the person in the photo. Interested in links to finished projects (services) or materials on how to do it yourself. At the same time, it is desirable that changes be minimally noticeable for a person.
I found some materials but nothing concrete so far, for example here:
https://www.secnews.ru/foreign/23902.htm#axzz6QzWp7twd (the service described in the news is not available)
https:// vc.ru/future/90049-facebook-razrabotala-te... (the video is good, of course, but I'm interested in the photo)

Answer 1

fawkes algorithm allows you to protect photos from recognition

Answer 2

Someone Malevich came up with the right filter back in 1915, here it is

More seriously. Hypothetically, you could write a program that contains both a face recognizer and a photo mutilator, run it iteratively, and see when the recognizer breaks.
It is very possible that the neural network of a person looking at such a photo will break earlier.

Answer 3

The topic is interesting. It is located at the intersection of the development of image recognition systems and general human medicine.
To protect - most likely it is possible. Only after that the photo will become inaccessible to human perception. Simple distortion filters (distort), barrel, pillow, torsion (twirl). Just some warps that change the proportions of the face, for example, stretching the eyes and so on - can break the system of searches by coincidence.
But if you decide to play cat and mouse with the state, then nothing will come of it. You will still get caught. And distorted photos in your social network will simply cause surprise and suspicion.
Maybe if you don't want to be searched for, then it's probably better not to publish anything at all.
Therefore, the disclaimer - the interest of this topic - is exclusively academic.