AML.T0042

MITRE ATLAS / Verify Attack

Adversaries confirm their attack works before deploying it, using an inference API or an offline copy of the target model to check that the crafted input produces the intended effect.

Think of it likeTesting the key in the lock once, quietly, well before the night you intend to use it.

In plain English

A failed attempt is a detection event. Verifying against a copy or a few careful queries means the attack that finally runs against production works the first time.