MITRE

Stefan Coetzeecreated 2026-05-12updated 2026-06-082 min readMOC

Vault note, not reviewed against the source. Written in the knowledge vault on 2026-05-12 by models working with Stefan Coetzee and published as it stands, with private addresses, e-mail addresses and an employer name redacted. Check claims against the primary source before relying on them.

MITRE Adversary Frameworks Cluster

Map of MITRE's adversary-and-defense knowledge bases: ATT&CK (adversary tactics, techniques, procedures), D3FEND (defensive countermeasures), ATLAS (adversarial AI threats). De facto operational threat-intelligence references. Reference cluster for threat modeling, incident response, security control selection, AI-specific adversarial threats. Free, openly licensed, community-maintained.

Anchors

position: current view, dated, revisable
anchors: MITRE Corporation, contributing community

Provenance

MITRE Corporation — US federally funded research and development center (FFRDC). Operates multiple FFRDCs for US government.
ATT&CK — first published 2013. Adversary tactics, techniques, procedures (TTPs) framework.
D3FEND — published 2021. Defensive countermeasures framework. Funded by NSA.
ATLAS — published 2020 (initial), 2021-2024 expanded. Adversarial Threat Landscape for AI Systems.

What MITRE frameworks provide

Three complementary knowledge bases:

ATT&CK — what adversaries do (offensive perspective).
D3FEND — what defenders do (defensive perspective).
ATLAS — what adversaries do against AI systems specifically.

Together provide the operational threat-intelligence and defense-mapping foundation for cybersecurity programs.

ATT&CK overview

Adversary behaviors organized as:

Tactics — adversary goals (e.g., Initial Access, Persistence, Lateral Movement, Exfiltration).
Techniques — methods to achieve tactics (e.g., Phishing, Valid Accounts, PowerShell, Web Service).
Sub-techniques — specific variations.
Procedures — concrete implementations observed in the wild.

Multiple matrices: Enterprise, Mobile, ICS, Cloud variants.

Detail in MITRE ATT&CK.

D3FEND overview

Defensive countermeasures organized as:

Tactics — defensive goals (e.g., Harden, Detect, Isolate, Deceive, Evict).
Techniques — defensive methods (e.g., Application Hardening, File Analysis, Network Isolation).

Mapped to ATT&CK techniques — for each adversary technique, possible defensive countermeasures.

Detail in MITRE D3FEND.

ATLAS overview

AI-specific adversary tactics:

Adapted ATT&CK structure for AI / ML systems.
Tactics include ML Model Access, ML Attack Staging, Initial Access, Defense Evasion, Discovery, etc.
Techniques include Adversarial ML Attack, Data Poisoning, Model Theft, ML Supply Chain Compromise, Prompt Injection.

Detail in MITRE ATLAS.

Why this matters for SRE and AI-agent work

Threat modeling organized around adversary behavior produces realistic threat models.
Control selection via D3FEND mapping to ATT&CK techniques.
AI-specific threats captured in ATLAS — most operational taxonomy for AI adversaries.
Incident response uses ATT&CK techniques as classification taxonomy.
Red team / blue team coordination uses common ATT&CK language.
Security product evaluation vendors map products to ATT&CK; comparison facilitated.

OWASP LLM Top 10 — application-level AI threat taxonomy; complementary to ATLAS
NIST AI RMF — threat-informed risk management
ISO 27001 — A.5.7 threat intelligence input source
ISO 42001 — AI-system threat input

Conventions

Atoms named MITRE <Framework>.md
Specific technique IDs (e.g., T1566 Phishing, AML.T0051 Prompt Injection) cited explicitly
Cross-link to OWASP LLM Top 10 / NIST AI RMF / ISO 27001