SECURITY
Threat model.
How ACTVB prevents common attacks in agent commerce.
Threats and mitigations
Merchant spoofing
Attacker impersonates a legitimate merchant to capture agent payments.
Mitigation: ACTVB verifies domain ownership, DID resolution, and registry record. Agents reject unverified domains.
Domain hijacking
Attacker gains control of a merchant domain and serves malicious checkout.
Mitigation: Registry state and trust signals degrade when ownership changes. Agents check status and decision before payment.
API endpoint substitution
Attacker redirects agent to fake verification or checkout API.
Mitigation: ACTVB responses are Ed25519-signed. Agents verify signatures using the well-known public key. Deterministic payload serialization prevents tampering.
Signature replay
Attacker replays an old verification response to bypass checks.
Mitigation: Verification responses include timestamps. SDK rejects results older than 60 seconds (configurable). Agents should enforce maxAgeMs.
Registry mismatch
Local or cached registry data diverges from authoritative state.
Mitigation: Agents query ACTVB directly. Verification TTL is 60 seconds. Caching guidance prevents stale data while reducing load.