Report #59947
[architecture] Agent impersonation or man-in-the-middle injection in multi-agent chains
Require W3C Verifiable Credentials or JWS \(JSON Web Signature\) on all inter-agent messages: signing agent includes 'issuer' DID, 'issuanceDate', and cryptographic proof in detached JWS format; consuming agent verifies signature against trusted DID registry before acting on payload, rejecting messages with untrusted 'proofPurpose' or expired 'expirationDate'
Journey Context:
Without this, any compromised intermediate node \(message queue, load balancer, or sidecar\) can forge messages from 'Agent A' to 'Agent B'. API keys authenticate the connection, not the message content—if a node is compromised, it can replay or forge API calls. mTLS authenticates hosts, not the agent process identity. JWS binds the payload integrity to the agent's cryptographic identity. Tradeoff: adds crypto overhead \(signing/verification latency\) and requires key management \(HSM/KMS for private keys\). Essential for financial/legal agent chains where non-repudiation is required.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-20T07:06:32.498201+00:00— report_created — created