Mitigation

AML.M0002Passive AI Output Obfuscation

What it is

Decreasing the fidelity of model outputs provided to the end user can reduce an adversary's ability to extract information about the model and optimize attacks for the model.

References

  1. https://atlas.mitre.org/mitigations/AML.M0002

Related by meaning· 6

Nearest entities by semantic similarity across the cs-graph corpus.

ATLAS mitigation
Encrypt Sensitive Information
ATLAS mitigation
Model Hardening
ATLAS mitigation
Limit Public Release of Information
ATLAS mitigation
Limit Model Artifact Release
ATLAS mitigation
Adversarial Input Detection
ATLAS
LLM Prompt Obfuscation
Sourced from MITRE ATLAS — Adversarial Threat Landscape for AI Systems. Curated by Adam Lundqvist, SQUR.