Trust and responsibilities¶
Security rests on several distinct responsibilities. The issuer makes the statement; a signing system protects the signing operation; a verifier checks specified properties; the recipient decides whether to use the evidence. One party may perform several roles, but the roles and their permissions must remain distinguishable.
This page describes the reference model. It does not establish a custody arrangement, trust service or operational control for any deployment.
Responsibility boundaries¶
- Professional source and issuer: Responsibility in the model: Authorise the content, identify the subject and scope, protect access to signing and issue attributable corrections or status updates. What the role does not establish: That every signed statement is factually correct, or that a recipient has permission to rely on it.
- Signing or custody provider, where used: Responsibility in the model: Enforce the issuer's authorised signing instructions and protect the keys and approval records assigned to it. What the role does not establish: Professional competence or authority to change the issuer's statement.
- Trust-list publisher or trust framework, where used: Responsibility in the model: Maintain the identity, key, role and scope records it is responsible for, including changes and withdrawal. What the role does not establish: Universal recognition, custody of participants' private keys or the recipient's acceptance.
- Holder or presenting party: Responsibility in the model: Share the selected evidence under applicable permissions and preserve its signed relationships. What the role does not establish: That the presentation contains every relevant record or every claim that could be disclosed.
- Verifier operator: Responsibility in the model: Perform the declared checks with the required inputs and report their results, failures and limitations accurately. What the role does not establish: The truth of the underlying statement, legal reliance, insurance cover or the recipient's decision.
- Recipient: Responsibility in the model: Select trusted sources and policy, supply the intended asset and purpose, assess freshness and completeness, and record acceptance separately. What the role does not establish: That another recipient must reach the same decision.
- Specification and software publisher: Responsibility in the model: Define clear rules, identify versions, document implementation limits and correct errors. What the role does not establish: That every implementation or deployment follows those rules.
These responsibilities describe technical boundaries. Contracts, professional obligations and applicable law determine the parties' actual duties.
A reference flow¶
The diagram shows where evidence and recipient-controlled inputs meet. It is not a deployment diagram or a claim that a production service implements these controls.
flowchart TB
subgraph Issuer["<br>Issuer boundary"]
E["Approved statement"] --> S["Authorised signing"]
end
S --> C["Credential and disclosures"]
subgraph Recipient["<br>Recipient boundary"]
P["Trusted keys and recipient policy"] --> V["Technical checks"]
T["Trust and status information"] --> V
V --> R["Scoped technical result"]
R --> A["Recipient acceptance decision"]
end
C --> V
click E href "issuer-and-key-security.html#approve-the-content-that-will-be-signed"
click S href "issuer-and-key-security.html#key-lifecycle"
click P href "verifier-behaviour.html#inputs-the-recipient-controls"
click T href "verifier-behaviour.html#freshness-and-disclosure"
click V href "verifier-behaviour.html#a-result-with-separate-checks"
click R href "verifier-behaviour.html#a-result-with-separate-checks"
click A href "#evidence-time-and-acceptance"
class Issuer,Recipient rwGroup
class E,S,C,P,T,V,R,A rwOutline
class E,S,C,P,T,V,R,A rwEven
The recipient's expected asset, purpose and policy must not be silently replaced by values supplied by the presenter. An interface or agent may carry those inputs; it does not gain authority to weaken them.
Key ownership is not professional authority¶
A key can be mathematically valid without being trusted. A trusted key can identify an organisation without establishing that the organisation is authorised to make a particular professional statement. Both checks are needed.
A recipient therefore needs a basis for connecting a key to an issuer, and a separate basis for recognising that issuer's role and scope. A key embedded in a credential, a familiar display name or an HTTPS connection alone is insufficient. A trust framework may provide relevant identity or authority records; that does not mean it must generate, distribute or hold the participant's private signing key.
The relevant policy and trust records should identify whose assertions the recipient accepts, for which statement types and during which periods. Trust in a valuation source does not automatically extend to a condition source or to a status-list publisher.
Evidence, time and acceptance¶
The edition of a specification, the status of an individual credential and the recipient's acceptance decision describe different things. A draft specification can describe a synthetic credential; a valid credential can still be unsuitable for a particular recipient or purpose.
Each result should identify the evidence presented, the policy applied and the time of checking. Missing evidence, unavailable status information and an unsupported check must remain visible. If a required check cannot be completed, the result must not claim that requirement passed. This is a requirement for implementations, not a statement that the current reference tools enforce every part of it.
A property valuation and a property condition credential are signed separately. Checking one does not authenticate the other. Their subject, relationship and relevant dates must also be checked. Re-signing a record does not update its valuation or inspection date.
Risks that remain¶
- An authorised source can make a mistake or a false statement. Source attribution supports accountability; it is not proof of truth.
- Signing access or a trust publisher can be compromised. Recovery requires evidence of the affected scope and changes to trust, not just a new signature.
- Revocation and status updates can arrive late. Offline checking depends on the freshness and provenance of the available snapshot.
- Verification software, configuration or an interface can produce a misleading result. The recipient must be able to identify what was actually checked.
- Disclosure controls cannot retract information already shared. A signed record may still contain confidential or identifying information.
No signature or security profile removes these risks. The assurance and release requirements explain what evidence is needed to support a narrower, verifiable security claim.