AIDEFEND
Open knowledge base of 70+ defensive countermeasures mapped to MITRE ATLAS, MAESTRO, and OWASP.
What it does
An open-source knowledge base of 70+ defensive countermeasures for protecting AI/ML systems. Three switchable views: Tactics (aligned to MITRE D3FEND), Pillars (Data, Model, Infrastructure, Application), and Phases (Design through Incident Response). Each technique includes implementation strategies, code examples, and tool recommendations. CC-BY-4.0 licensed.
Security relevance
The defensive counterpart to MITRE ATLAS. While ATLAS maps how AI systems are attacked, AIDEFEND maps how to defend them. Techniques are explicitly mapped to known threats from MITRE ATLAS, MAESTRO, and OWASP LLM Top 10 — making it the most comprehensive defensive reference available. Includes a local MCP/REST API for programmatic access.
When to use it
Use as a primary reference when designing AI security controls. The three views serve different roles: Tactics for security architects, Pillars for ML engineers, Phases for DevSecOps teams. Search by technique ID, threat mapping, or keyword. No login required for the interactive web interface.
OWASP coverage
Risks addressed — mapped to both OWASP Top 10 standards. 10 in LLM, 5 in Agentic.
The raw record
What Yuntona stores. Single source of truth — fork it on GitHub.
name: AIDEFEND slug: aidefend type: Mixed category: AI Governance & Standards url: https://aidefend.net reviewed: 2026-04 added: 2026-04 updated: 2026-04 risks: llm: [LLM01, LLM02, LLM03, LLM04, LLM05, LLM06, LLM07, LLM08, LLM09, LLM10] asi: [ASI01, ASI02, ASI03, ASI04, ASI07] complexity: Guided Setup pricing: — audience: All lifecycle: [develop] tags: [Defense, Framework, Open Source, Threat Mapping]