Framework Mappings

The mappings field links threat models to industry security frameworks.

Overview

{
  "mappings": {
    "mitreAttack": [...],
    "mitreAtlas": [...],
    "owasp": [...],
    "stride": [...],
    "cwe": [...],
    "cvss": {...}
  }
}

MITRE ATT&CK

{
  "mitreAttack": [
    {
      "tacticId": "TA0001",
      "tacticName": "Initial Access",
      "techniqueId": "T1189",
      "techniqueName": "Drive-by Compromise",
      "description": "Malicious website exploits browser",
      "url": "https://attack.mitre.org/techniques/T1189"
    }
  ]
}

Field	Type	Required	Description
tacticId	string	Yes	Tactic ID (TA####)
tacticName	string	No	Tactic name
techniqueId	string	Yes	Technique ID (T####)
techniqueName	string	No	Technique name
description	string	No	Context
url	string	No	ATT&CK URL

MITRE ATLAS

{
  "mitreAtlas": [
    {
      "id": "AML.T0024",
      "name": "Prompt Injection",
      "description": "Crafted prompts manipulate AI behavior"
    }
  ]
}

Field	Type	Required	Description
id	string	Yes	ATLAS technique ID (AML.T####)
name	string	No	Technique name
description	string	No	Context
url	string	No	ATLAS URL

Alignment Failure Modes

Alignment failure modes classify why AI models behaved as they did — a layer that cyber (ATT&CK) and adversarial-ML (ATLAS) frameworks do not capture. Use them for incidents rooted in misaligned model behavior (e.g. reward hacking, goal drift, emergent coordination).

{
  "alignmentFailureModes": [
    {
      "id": "reward-hacking",
      "name": "Reward Hacking",
      "description": "Agent optimizes the reward signal via an unintended path rather than solving the task",
      "primary": true
    }
  ]
}

Field	Type	Required	Description
id	string	Yes	Stable slug (e.g. reward-hacking)
name	string	No	Human-readable failure-mode name
description	string	No	How the failure mode manifested
primary	bool	No	Marks the dominant failure mode for the incident

Section Summaries

mappings.summaries holds an optional short prose intro per classification section, keyed by section slug (cve, cwe, mitreAttack, mitreAtlas, owasp, stride, alignmentFailureModes). Renderers use it to introduce each mapping table.

{
  "summaries": {
    "alignmentFailureModes": "Why the collective behaved as it did, beyond the cyber and adversarial-ML views."
  }
}

OWASP

{
  "owasp": [
    {
      "category": "api",
      "id": "API2:2023",
      "name": "Broken Authentication",
      "description": "No rate limiting on authentication",
      "url": "https://owasp.org/API-Security/..."
    },
    {
      "category": "agentic",
      "id": "ASI02:2026",
      "name": "Tool Misuse & Exploitation",
      "description": "Attackers exploit agent's access to tools"
    }
  ]
}

Field	Type	Required	Description
category	string	Yes	web, api, llm, or agentic
id	string	Yes	OWASP ID
name	string	Yes	Vulnerability name
description	string	No	Context
url	string	No	OWASP URL

OWASP Categories

Category	Value	Year
Web Applications	web	2021
API Security	api	2023
LLM Applications	llm	2025
Agentic Applications	agentic	2026

Attack Step OWASP Mappings

Attack steps can include owaspIds and asiIds:

{
  "attacks": [
    {
      "step": 1,
      "from": "attacker",
      "to": "agent",
      "label": "Inject malicious prompt",
      "owaspIds": ["LLM01:2025"],
      "asiIds": ["ASI01:2026", "ASI02:2026"]
    }
  ]
}

STRIDE

{
  "stride": [
    {
      "category": "S",
      "name": "Spoofing",
      "description": "Attacker impersonates legitimate client"
    },
    {
      "category": "I",
      "name": "Information Disclosure",
      "description": "API keys exposed to attacker"
    }
  ]
}

Field	Type	Required	Description
category	string	Yes	S, T, R, I, D, or E
name	string	No	Category name
description	string	No	Context

CWE

{
  "cwe": [
    {
      "id": "CWE-346",
      "name": "Origin Validation Error",
      "description": "WebSocket accepts any origin",
      "url": "https://cwe.mitre.org/data/definitions/346.html"
    }
  ]
}

Field	Type	Required	Description
id	string	Yes	CWE ID (CWE-###)
name	string	No	Weakness name
description	string	No	Context
url	string	No	CWE URL

CVSS

{
  "cvss": {
    "version": "3.1",
    "vector": "CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:C/C:H/I:H/A:N",
    "score": 9.3,
    "severity": "Critical"
  }
}

Field	Type	Required	Description
version	string	Yes	CVSS version
vector	string	Yes	CVSS vector string
score	float	No	Calculated score
severity	string	No	Critical/High/Medium/Low

Complete Example

{
  "type": "attack-chain",
  "title": "WebSocket Localhost Takeover",
  "mappings": {
    "mitreAttack": [
      {
        "tacticId": "TA0001",
        "tacticName": "Initial Access",
        "techniqueId": "T1189",
        "techniqueName": "Drive-by Compromise"
      },
      {
        "tacticId": "TA0006",
        "tacticName": "Credential Access",
        "techniqueId": "T1110",
        "techniqueName": "Brute Force"
      }
    ],
    "mitreAtlas": [
      {
        "techniqueId": "AML.T0024",
        "techniqueName": "Prompt Injection"
      }
    ],
    "owasp": [
      {"category": "api", "id": "API2:2023", "name": "Broken Authentication"},
      {"category": "api", "id": "API4:2023", "name": "Unrestricted Resource Consumption"},
      {"category": "llm", "id": "LLM06:2025", "name": "Excessive Agency"},
      {"category": "agentic", "id": "ASI02:2026", "name": "Tool Misuse & Exploitation"}
    ],
    "stride": [
      {"category": "S", "name": "Spoofing"},
      {"category": "I", "name": "Information Disclosure"},
      {"category": "E", "name": "Elevation of Privilege"}
    ],
    "cwe": [
      {"id": "CWE-346", "name": "Origin Validation Error"},
      {"id": "CWE-307", "name": "Improper Restriction of Excessive Authentication Attempts"}
    ],
    "cvss": {
      "version": "3.1",
      "vector": "CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:C/C:H/I:H/A:N",
      "score": 9.3,
      "severity": "Critical"
    }
  }
}

STIX 2.1 Export

Framework mappings are preserved in STIX 2.1 export:

• MITRE ATT&CK → Attack Patterns with external references
• MITRE ATLAS → Attack Patterns with external references
• CWE → Vulnerability objects

tms generate model.json --stix -o model.stix.json