#threat-modeling

Adversarial Examples vs. Data Poisoning: Timing Is Everything

Adversarial examples attack a deployed model at inference; data poisoning attacks the model before it is deployed.