• The idea behind transferability attacks is pretty simple: that adversarial

Transferability attacks take advantage of the fact that adversarial examples created for one model are often effective against other models. An attacker can craft adversarial examples on a surrogate model and then use them against the target model.