Directory context | Adversarial Machine Learning

MITRE ATLAS

Knowledge base of adversary tactics and techniques against AI systems; successor to the deprecated Adversarial ML Threat Matrix

Direct answer

What is MITRE ATLAS?

MITRE ATLAS is included in the Awesome MLSecOps Adversarial Machine Learning directory. The community-maintained README describes it as: “Knowledge base of adversary tactics and techniques against AI systems; successor to the deprecated Adversarial ML Threat Matrix.” Its MLSecOps relevance is the controlled evaluation of evasion, poisoning, extraction, inversion, privacy attacks, or robustness under an explicit attacker model. The linked first-party source is published at atlas.mitre.org. A technical review should test the project's documented evidence across four criteria: Threat-model fit, Modality and framework support, Adaptive attack evaluation, and Experiment reproducibility. Compare that evidence with the intended architecture and threat model. Catalog inclusion establishes relevance to this security category; it is not a certification, comparative ranking, or endorsement. Confirm current capabilities, maintenance, licensing, limitations, and deployment assumptions in the first-party documentation before adoption.

Knowledge base of adversary tactics and techniques against AI systems; successor to the deprecated Adversarial ML Threat Matrix

Neutral catalog description synchronized from the Awesome MLSecOps README

Before adoption

What should teams verify about MITRE ATLAS?

Answer these questions from current first-party documentation and testing evidence rather than relying on the directory listing alone.

  1. 01

    Which attacker capabilities, access levels, and success conditions are modeled?

  2. 02

    Which data modalities, frameworks, and model interfaces are supported?

  3. 03

    Do defense evaluations include adaptive attackers and meaningful baselines?

  4. 04

    Can experiments be reproduced from versioned configurations, datasets, and metrics?