Adversarial Examples — AI Dictionary

An input crafted with a small, often imperceptible perturbation that causes a model to produce a wrong or attacker-chosen output.