Implementation of adversarial attacks and defences (github.com)

Note that this pertains to exploiting ML systems so that you control the inference (output) of the system.