I think what you are looking for is called "Adversarial Design".
Article: https://news.ycombinator.com/item?id=21482372 Archive: https://archive.today/HuanH
Defcon: Tutorial on hacking Facial Recognition systems. https://www.youtube.com/watch?v=b-lkYtprgVw
[PDF] Adversarial Examples that Fool both Computer Vision and Time-Limited Humans
(post is archived)