Agent Impersonation
Security & IdentityDescription
Agents or adversaries impersonate legitimate agent identities, enabling trust exploitation, privilege inheritance, and policy bypass. Absence of strong agent authentication enables systematic identity spoofing in multi-agent systems.
Low-privilege agent spoofs the orchestrator's identity to issue instructions to a privileged sub-agent, bypassing authorisation checks.
Adversarial input manipulation has been rigorously demonstrated in controlled research across image classifiers, NLP models, and autonomous systems, including physical-world attacks. Confirmed large-scale production exploitation causing documented harm at scale has not yet been publicly verified.
Primary mitigations
- Cryptographic agent identity (PKI)
- agent certificate management
- per-message authentication in A2A communication
- identity verification at trust boundaries.
Detection signals
Agent Identity Authenticity Score; impersonation detection rate; failed identity verification events.