Repository navigation
Question: signer-bound handoff for an exported Annex IV evidence dossier? #4541
|
Hi — I’m working on BoundaryAttest, a small open-source experiment around portable evidence when an artifact crosses from one trust domain to another: https://github.com/cullenmeyers/BoundaryAttest I read the current AGT evidence/export path before reaching out, particularly the Annex IV exporter/evidence pipeline and the Merkle audit design. My understanding is:
I would not want BoundaryAttest to replace or reinterpret any of that native evidence. The boundary I’m interested in is later: At that point there seem to be two different questions:
I’m trying to learn whether #2 is actually useful in a real workflow or whether ordinary artifact signing / existing AGT mechanisms already make a separate receipt redundant. BoundaryAttest’s claim would be intentionally narrow:
It would not prove that:
Would anyone maintaining this area be willing to try the generic CLI once against a synthetic/non-sensitive representative For example: npm install boundaryattest@0.1.1
npx boundaryattest keygen
npx boundaryattest attest ./annex-iv-report.md \
--action-type agt.annex_iv_handoff \
--private-key ./boundaryattest-private.pem \
--output ./annex-iv-report.receipt.jsonThen a separate recipient would receive: with the expected public key obtained through a separately trusted channel, and run: npx boundaryattest verify-artifact \
./annex-iv-report.md \
./annex-iv-report.receipt.json \
--public-key ./boundaryattest-public.pemNo AGT code changes, no adapter and no production key required. What I’m most interested in is the result, not getting BoundaryAttest integrated. If the reaction is:
that is useful evidence and I would treat BoundaryAttest as redundant at this boundary. If the structured handoff claim is useful beyond ordinary file signing, I’d like to understand specifically what additional semantics the recipient cares about. No implementation request — just one small outsider-use test if the boundary is realistic. |
Replies: 2 comments 3 replies
|
I can't run the CLI test for you, but I can answer the part you said you care about most: what a recipient needs beyond plain file signing. There is also one thing in the current exporter that matters before any outer signature does. 1. The manifest does not yet bind the report to the evidence underneath it In
So a signature over 2. What the recipient cares about, from the regulation text The handoffs in the Act are to national competent authorities and notified bodies. Article 11(1) says the documentation exists "to provide national competent authorities and notified bodies with the necessary information". These are the semantics they will need that a bare file signature does not carry:
3. So is a separate handoff receipt redundant? For integrity of the bytes, yes: Sigstore or minisign already does that. What they do not give you is the table above: version lineage, provider-accountable identity, long-term verifiability, and request context. That is where a structured handoff statement earns its place. Two notes on the design:
The non-claims you listed (completeness, truthfulness, compliance, acceptance) are right, and they belong in the statement itself so that no recipient reads more into the signature than it says. |
|
Glad it helped, and thanks for checking John B Joseph, Co-founder & CPTO, Purplecop Security |
I can't run the CLI test for you, but I can answer the part you said you care about most: what a recipient needs beyond plain file signing. There is also one thing in the current exporter that matters before any outer signature does.
1. The manifest does not yet bind the report to the evidence underneath it
In
evidence_pipeline.py,EvidenceSourcedeclaressha256: str = ""(L66), but:to_dict()(L68-74) does not emit it, sosave_manifest()(L130-147) writes onlysource_type,path,collected_atandrecord_countfor each source.So a signature over
annex-iv-report.md, from BoundaryAttest, Sigstor…