Training data poisoning
Pre-training, fine-tuning or embedding data tampered to corrupt behaviour
categoryAI/ML and emerging attacksseverityHighlabnot yet measured
Classification
15 values: 9 reviewed, 5 imported, 1 derived. What do these mean?
Exploitability
| Attack vector | AI model manipulation |
|---|---|
| Reach (CVSS) | Network |
| Attack complexity | High |
| Privileges required | Low |
| User interaction | None |
| Attacker skill required | High |
| Resources required | Medium |
| Likelihood / prevalence | Medium |
| Maturity | Emerging |
Impact & scope
| Severity | High |
|---|---|
| Impact | Integrity |
| Platform | AI System |
| Environment | AI system |
| Target asset class | ML model / agent |
| Kill-chain phase (tactic) | Persistence |
Detection & response
Mitigated by 7
- AML.M0001 atlas-mitigation
- AML.M0005 atlas-mitigation
- AML.M0007 atlas-mitigation
- AML.M0008 atlas-mitigation
- AML.M0023 atlas-mitigation
- AML.M0025 atlas-mitigation
- AML.M0035 atlas-mitigation
Framework references
- Primary source: https://genai.owasp.org/llm-top-10/
- owasp-llm: LLM04:2025 (primary)
- atlas: AML.T0020
Evidence & sources
- ev-0001 (Cyber Attack Taxonomy) — Name, definition and Example/ID taken verbatim from the 'AI/ML and emerging attacks' table, line 615 of Cyber-Attack-Taxonomy.md