Adversarial Examples

Inputs with small perturbations that cause models to produce incorrect outputs.