Back to MITRE ATLAS
MITRE ATLAS · AML.M0014
Adversarial training
“Train models on adversarial examples or with robust optimization to reduce sensitivity to crafted inputs.”
ATLAS mitigation AML.M0014
model
Type
mitigation
Parent
—
Sub-items
0
Cross-framework relationships
Lattice uses an explicit, four-type vocabulary and refuses ‘equivalent.’ Each relationship below carries a rationale and an editorial reviewer credit.
informs
1- Evade ML modelMITRE ATLAS · AML.T0015
Evasion is mitigated in part by adversarial training within ATLAS's mitigation catalog.
Reviewer: Editorial: K. Park · Reviewed 2026-05-02
Version: Pinned 2026.Q1Last reviewed: 2026-05-02