RFC-032: Post-Hoc Evaluation Loop — 13. Security Considerations
AIGP Specification › RFC-032: Post-Hoc Evaluation Loop › 13. Security Considerations
← 12. Burden of the Inquiry · Section index · 14. Relationship to Existing RFCs →
13. Security Considerations
| Concern | Mitigation |
|---|---|
| Retroactive ANTICIPATE modification | D-DNA signing + seal-on-first-execution + rejection of modifications after seal |
| Self-evaluation (agent grades itself) | Structural prohibition: grader_id ≠ agent_id, enforced at protocol level |
| Grader manipulation | MODEL graders operate in isolated execution context; input is ANTICIPATE + evidence only |
| Verdict tampering | VERIFY messages are D-DNA signed; governance actions are logged immutably |
| Denial-of-service via excessive ANTICIPATE | Rate limiting on ANTICIPATE submissions per agent per session |
| Gaming the grader | Calibration program detects systematic grader bias; HUMAN baseline prevents drift |
← 12. Burden of the Inquiry · Section index · 14. Relationship to Existing RFCs →