Context
Assay currently proves tamper-evidence: receipts were not modified after creation. It does not prove source-attestation: the system could lie when creating receipts. This is a known, documented limitation (PACK_SUMMARY.md "What This Does NOT Prove").
Common question from launch: "if the system is malicious it can emit fake receipts."
Layers to explore
- TEE-backed signing -- Sign receipts inside a Trusted Execution Environment so the key is hardware-protected
- External witness -- Third-party timestamp authority (RFC 3161) to anchor pack creation time
- Transparency log -- Append-only public log (assay-ledger is the start) with inclusion proofs
- Cross-system corroboration -- Multiple independent systems producing receipts for the same interaction
- Hardware attestation -- TPM/SGX quotes proving the signing environment is unmodified
Priority
Post-launch. The current trust boundary (tamper-evidence + signed packs) is sufficient for v1. Source attestation is the v2 trust layer.
References