Adversarial ML & LLM Application Security
188 AI security use cases,
from model to application.
Every technique in MITRE ATLAS (178, covering the full attack lifecycle from reconnaissance through model theft and agentic exfiltration) plus the 10 categories in OWASP's Top 10 for LLM Applications (2025) — indexed by tactic, maturity, and an inferred log source since neither framework publishes a native data-source taxonomy the way MITRE ATT&CK does.
"
Don't boil the ocean.92 of 178 ATLAS techniques are only "Demonstrated," not "Realized" in the wild — weight detection investment toward the 68 Realized techniques and the 10 OWASP categories your app stack actually exposes.
Browse the library
filter by framework, search by name, tactic, or log source
loading library…