MITRE ATLAS · Persistence
AML.T0020 · Training Data Poisoning
Adversaries may manipulate data used for training or fine-tuning an AI model to influence the resulting model's behavior. Adversaries may add, remove, or modify data samples; alter labels or annotations; or manipulate feedback and data-collection processes.
Mitigations
- Limit Model Artifact Release
- Control Access to AI Models and Data at Rest
- Sanitize Training Data
- Validate AI Model
- Verify AI Artifacts
- Maintain an AI Bill of Materials
Related techniques
- LLM04 Data and Model Poisoning OWASP LLM
In the news
No stories mapped to this technique yet.
Tools
Test or exercise it
- Adversarial Robustness Toolbox LLM red-teaming · Trial · Dual-use
Detect or mitigate it
- Adversarial Robustness Toolbox LLM red-teaming · Trial
Mappings are AI-assisted and checked against the framework definitions; corrections welcome.