DEV Community

Cover image for What a SCITT Receipt doesn't tell your verifier
VeritasChain Standards Organization (VSO)
VeritasChain Standards Organization (VSO)

Posted on Fully Autonomous

What a SCITT Receipt doesn't tell your verifier

If you verify SCITT Receipts, here is a gap worth knowing about. We just recorded it as an open issue in the SCITT profile for the VeritasChain Protocol (VCP), in draft-kamimura-scitt-vcp-06. The draft does not solve it.

What an RFC 9942 inclusion Receipt actually signs

For the RFC9162_SHA256 data structure (RFC 9942 §5.2.1):

  • The payload is the Merkle tree root.
  • The protected header carries the signature algorithm (label 1) and the data structure identifier (vds, label 395).
  • The unprotected header carries the proof (vdp, label 396): tree size, leaf index and inclusion path.

Verification is two steps: apply the inclusion path to the bytes of a candidate entry to recompute the root, then check the signature over that root.

Three consequences:

  1. You need the entry rule. RFC 9943 §5.1.3 leaves the data structure to the Transparency Service, and RFC 9942 §4 notes that leaf construction can differ. Is the leaf the statement bytes? A digest of something? The Receipt does not say.
  2. You need the right key. The Receipt is signed by the service. Which key you trust for which service is outside the Receipt.
  3. Tree size and leaf index are not signed. They are authenticated only through root recomputation. If your argument depends on where a statement sits in the log, or how big the log was, decide whether that is enough or whether you also need a signed tree head or a consistency proof under a trusted log key.

The hackathon report presented to the SCITT Working Group at IETF 127 reached the same three points from practice: Receipts verified across operators only with rules agreed out of band, namely what the leaf is, which key signs and what the tree size was. That was other teams' work on their own drafts, not an implementation of VCP.

Why this matters for an audit-trail profile

VCP's omission reasoning relies on counts and sequence: sequence numbers, anchored batch roots that bind an event count and endpoints, and heartbeat events. The profile already says that SCITT inclusion proofs do not by themselves establish completeness. If part of such an argument is carried by a Transparency Service's log, the three questions above decide what a verifier can actually conclude.

What changed in this round

  • draft-kamimura-scitt-vcp-06: Open Issue 6 (above); Open Issues 1 to 5 (Merkle construction, AnchorRecord signature scope, integer timestamps, EventHash notation, Security object fields) now point to proposed VCP change proposals, all unadopted; Implementation Status adds VSO's own pre-release VCP v1.2 library (Python, TypeScript, MQL5), which does not read COSE or SCITT envelopes.
  • draft-kamimura-vap-framework-03: maps the draft's three explanatory groups to the five mandatory layers of VAP v1.2. No requirement changed.

What this does not do

No requirement, field or algorithm changed, and no open issue is resolved. These are individual Internet-Drafts, not adopted by any IETF Working Group and not endorsed by the IETF. VAP and VCP work after the fact: they make records auditable and attributable, and do not block anything. VAP v1.2 §1.6: conformance "does not warrant the correctness, fairness, or safety of the underlying AI decisions — only the integrity, completeness (at anchor granularity), and attributability of their records".

Disclosure: zero external implementations of VAP or any profile, and zero Evidence Packs accepted in any proceeding (count dated 25 October 2026, reviewed 11 November 2026). A previously published paid-contract figure was withdrawn on 24 October 2026; no contract, customer or revenue figure is claimed. Prepared with an AI assistant (Claude, by Anthropic); no independent review. VSO is operated by VeritasChain Co., Ltd.

Feedback on [email protected] is welcome, especially if you already pin entry rules and tree heads in your verifier.

Top comments (0)