MITRE ATLAS is a knowledge base of adversary tactics and techniques against machine learning systems. It is modelled on the MITRE ATT&CK framework and curated by MITRE.
Semantic Classification
Content
- MITRE ATLAS documents how attackers can target machine learning systems, organising real-world techniques such as data poisoning, model evasion, and model theft into a structured matrix. It draws on observed incidents and published research.
- Security teams use ATLAS to reason about threats to AI systems and to plan defences and evaluations. Its structure mirrors ATT&CK so that practitioners familiar with that framework can apply similar thinking to machine learning.