Published

Projection / Disclosure Policy

vr.wm-xct-003 · wm-xct-003-projection-disclosure-policy

Define, in a storage- and interface-neutral way, the shape data is permitted to leave in - which elements, after which transformations, over which records, at which grain - so one reusable shape specification can be bound to many contracts and audiences and can be reproduced and explained afterwards.

World Models Cross-cutting context XCT.DSC

Bundle → Layer → Finding → Questions Classifiers What it is Why it exists Distinguishing features What robots and AI may and may not do Moral aspects Owners Links to other meta-models Identity and identifiers Direct properties Recognition Capabilities and actions Hazards and failure modes Standards and interfaces Context of use Sources

Bundle → Layer → Finding → Questions Filled

6 bundles · 14 layers · 26 findings · 103 questions

Shape Semantics What an output shape actually is: which elements leave, how each is transformed, which records are in scope, at what grain, and how two shapes compare and combine.

Element Selection

Which elements of a source structure may leave, expressed as sets over a declared default, and how those elements are addressed.

Selection Sets and Default Treatment

Include and exclude sets over source elements, the default treatment applied to everything the sets do not name, and whether the shape is closed against elements it has never seen.

What treatment applies to a source element that neither the include nor the exclude set names? constraint
When an include path and an exclude path both match one element, which prevails and on what rule? decision
Which elements must survive selection for the output to remain usable, and what happens when the policy would remove them? requirement
How is an element treated that was added to the source after this shape was authored? exception

Path Expression and Element Addressing

How a shape addresses source elements across media, which grammar it uses, and what keeps an address resolvable when the source moves or repeats elements.

Which path grammar and grammar version does this shape use to address source elements? definition
How is an element located when the source is not a JSON tree, such as a tabular column, an RDF triple, a binary region or a span of free text? interoperability
How does an address stay resolvable when the source element is renamed, moved or occurs many times? identity
Which path constructs are forbidden because their result is non-deterministic or unbounded? constraint

Expansion, include, and hierarchical node selection

Output shape is a graph, not only a column list. OData $expand includes related resources inline and may nest $select; context URLs describe projected+expanded entities. GraphQL selection sets are hierarchical down to scalars. FHIR _include/_revinclude add referenced resources; they cannot mix with _summary=text. JSON:API compound documents require full linkage except where sparse fieldsets exclude relationship fields. XACML Multiple Decision Profile interprets a multiple:content-selector XPath as one decision per node. This model records maximum expand depth, allowed navigation paths, and whether related resources inherit the parent policy or a nested policy.

Which navigation/relationship paths may be expanded, and to what maximum depth? constraint
Does each expanded type use this policy, a nested projection-policy reference, or inherit parent treatments? relationship
For hierarchical XML/JSON resources, is a content-selector (XPath/JSONPath) used to name the nodes whose shape is governed, analogous to XACML multiple:content-selector? identity

Element Treatment

The transformation applied to a selected element before it leaves, its parameters, and whether it can be undone.

Treatment Technique and Parameters

The de-identification or shaping technique applied to an element, named from a governed vocabulary and parameterised well enough that the transformed value is reproducible.

Which technique is applied to this element and from which governed technique vocabulary is its name drawn? classification
Which parameters fully determine the transformed value, and may those parameters themselves be disclosed? measurement
Is the transformation deterministic across separate requests, and what does that imply for linkability? quality
Which technique-and-element combinations are refused outright regardless of who is asking? privacy

Reversibility and Re-identification Coding

Whether a treatment can be undone, by whom under which separate authority, and how a pseudonym or re-identification code namespace is referenced without this model holding a key.

Is this treatment reversible, and which party could reverse it under which separate instrument? authority
Where is the pseudonym or code namespace defined, and is it shared with any other shape? relationship
What must never be embedded in or derivable from a disclosed code or token? security
How is the irreversibility claim evidenced rather than merely asserted? evidence

Record Scope and Grain

Which records leave and at what level of summarisation, as distinct from which fields leave.

Population and Record Scope

The predicate deciding which records or instances appear in an output at all, and the disclosure created by membership in that set irrespective of field values.

Which predicate decides the set of records in an output, and over which attributes is it stated? composition
Does membership in the released record set by itself reveal a sensitive fact about a subject? privacy
What minimum and maximum record counts apply to an output, and which model owns those floors? constraint
How is record scope stated when the source is a stream or an unbounded collection? temporal

Aggregation Grain Declaration

For summary-only shapes, the dimensions retained, the measures computed and the period, with cohort floors and privacy budgets referenced rather than defined.

Which dimensions are kept in the summary and at what level of each hierarchy? definition
Which measures are computed, and with what rounding or perturbation applied to each? measurement
Over which period is the summary computed, and may a recipient recombine periods? temporal
Which privacy floor or budget instrument governs this grain, and where is it maintained? relationship

Leaving-shape grain class

A projection policy names exactly one primary grain for what may leave: (a) record-level subset — selected paths of authorised instances; (b) named summary — a predefined reduced view such as FHIR _summary=true/text/data; (c) aggregate-only — statistical functions over a grouping, with no microdata. ISO/IEC 20889 classifies aggregation and sampling as statistical de-identification tools; NIST SP 800-188 distinguishes publishing de-identified microdata, synthetic data, query interfaces, and protected enclaves. FHIR defines discrete summary modes including count-only.

What grain may leave under this policy: record-level subset, named summary, aggregate-only, or a documented hybrid? classification
If grain is summary, which named summary mode applies (for example FHIR true/text/data/count, or a locally named card)? classification
Are instance-level records permitted to leave at all, or only derived statistics? constraint
Which NIST SP 800-188 data-sharing model does this grain align to: publish de-identified data, synthetic data, query interface, or protected enclave — as alignment, not conformance? interoperability

Shape Algebra

The order relation on shapes and the operation that merges two applicable shapes into one no wider than either.

Narrowing Order and Least-Disclosure Combination

When one shape is no wider than another, how two applicable shapes combine into a single least-disclosure shape, and why this is distinct from deciding which policy takes precedence.

Under what test is one shape judged to disclose no more than another? validation
How are two applicable shapes merged, and is the result guaranteed no wider than either input? composition
Which merges are undefined or unsafe and must be refused rather than approximated? exception
How does merging two shapes differ from deciding which of two policies prevails? relationship

Compiled Output Template The concrete, verifiable shape a consumer actually receives, its derivation from pinned inputs, its encodings and the disclosure that survives beyond its stated values.

Template Derivation

How a template is produced from pinned inputs, identified, and kept aligned with a moving source schema.

Compiled Template and Fingerprint

The derived output shape a consumer receives, produced from one policy version against one source schema version and digested so any party can verify what was served.

Which inputs must be pinned for a template to be reproducible byte for byte? provenance
Over which canonical form is the template fingerprint computed, and with which algorithm? identity
Which part of a template may be shown to a prospective recipient before any authorization exists? access
When must a template be recompiled instead of reused? process

Source Schema Binding and Drift

The dependency of a shape on a governed source schema version, how unresolved paths and newly appearing elements are detected, and who remediates.

To which source schema or data contract version is this shape bound, and how strictly? relationship
Which selection paths currently fail to resolve against the bound schema version? validation
What happens when the source adds an element this shape has never classified? exception
How is drift detected, how often is the check run, and when was it last run? process

Disclosure Surface

How the shape is rendered into concrete encodings, and what a recipient can still infer beyond the stated values.

Encoding and Media Profile

Which encodings a shape is approved to be served in, whether semantics are preserved in each, and which encoding choices are themselves disclosure decisions.

In which encodings is this shape approved to be served, and is the disclosed content identical in each? interoperability
Does the encoding let a recipient distinguish an omitted element from one present but masked or null? definition
Which encoding-level metadata such as field order, annotations, comments or provenance headers leaves with the output? privacy
How is a holder-selected disclosure encoding constrained by an issuer-fixed shape? constraint

Residual Disclosure and Side Channels

What a recipient can still infer from a served output beyond its values: existence of a response, record counts, ordering, structural shape and the difference between error, refusal and empty result.

Which facts does a non-empty response reveal even when every value in it is masked? privacy
Does the count, ordering or structural padding of a response leak an attribute? quality
Do error, refusal and empty-result responses differ in a way that discloses the withheld value? exception
Which of these residual channels are accepted as tolerable, and on whose authority? decision

Binding and Applicability Where a shape applies: the targets and audiences it is attached to, the conditions under which it is the applicable shape, and the class and regime constraints that bound what is lawful.

Binding Attachment

The link record attaching a shape to a target and audience, and the declarative conditions for its applicability.

Binding Target, Audience and Effective Window

The attachment of one shape to a contract or grant, object type or audience class, with the window in which it is effective and the authority the binding presupposes.

What is the target of this binding, and which model owns that target? relationship
Which recipient class does the binding serve, and how is that class defined without copying party records? classification
From when until when is the binding effective, and what governs an output already served under it? temporal
Which authority must exist for this binding to be publishable at all, and what happens if it is absent? authority

Applicability Conditions

The declarative conditions under which this shape is the applicable one, stated so an external evaluator can decide without this model performing or recording any decision.

Which condition operands does this shape use, and from which governed operand vocabulary are they drawn? definition
Which attribute values must an evaluator be able to obtain for a condition to be decidable? interoperability
Where does declaring a condition end and deciding it begin? process
Which conditions are mandatory for applicability and which are advisory? constraint

Class and Regime Constraints

Externally imposed lower bounds on treatment: governed sensitivity classes and legally prescribed shapes.

Class-to-Treatment Minimum Matrix

The governed table mapping an externally assigned sensitivity class to the weakest treatment a shape may apply, so that shapes can be checked mechanically rather than by reviewer opinion.

For each referenced sensitivity class, what is the weakest treatment a shape may apply? requirement
Which classification scheme and version does the matrix key on, and how are keys resolved? provenance
How is a shape checked against the matrix, and what constitutes a violation? validation
What happens to existing shapes when an element is reclassified upward? event

Regime-Mandated and Regime-Forbidden Shapes

Shapes that a legal or sectoral regime prescribes or forbids outright, recorded as named, cited constraints with their territorial and sectoral reach, never as restated law.

Which regime prescribes a fixed output shape for this data, and what exactly does it prescribe? requirement
In which territories and sectors does that prescription bind, and what shape applies elsewhere? spatial
When two regimes prescribe incompatible shapes for one recipient, what is recorded and who resolves it? exception
How is a regime claim evidenced so it can be re-checked when the law changes? evidence

Assurance and Evidence Why a shape is defensible: the identifiability role of each element, the assurance method claimed, and the effect of serving several shapes over the same population.

Identifiability Declaration

The role each element plays in re-identification and the basis on which the resulting shape is claimed adequate.

Element Identifiability Roles

Whether each selected element is a direct identifier, a quasi-identifier, a sensitive attribute or non-identifying, and the context assumptions behind that assignment.

What identifiability role is assigned to each selected element, and from which terminology? classification
Which auxiliary datasets and adversary capabilities were assumed when assigning quasi-identifier status? provenance
How does an element's role raise the minimum treatment required for it? constraint
Which elements are deliberately left unassigned, and why is that defensible? quality

Assurance Method and Evidence Pointer

The basis on which the shape is claimed adequate - a rule-based method, an expert determination, or a formal privacy guarantee - held as a claim plus a pointer to evidence maintained elsewhere.

Which assurance method is claimed for this shape, and against which standard? decision
Who made the determination, when, and for how long does it remain valid? ownership
Where is the supporting evidence held, and what does this model deliberately not hold? evidence
What invalidates the claim, and how must the shape behave while it is invalid? state

Cross-Release Effects

Disclosure that arises from the set of shapes served rather than from any single shape.

Release-Set Linkability

The widening of disclosure when several shapes are served over the same population to the same or colluding recipients, through shared join keys, repeated queries or overlapping grains.

Which other shapes may be served over the same subjects to the same recipient class? relationship
Which elements act as join keys across shapes, and how is that overlap controlled? constraint
How many times may this shape be served over one population before the combined disclosure exceeds the claim? measurement
How is a change in the surrounding release set detected after this shape was approved? event

Interoperability and Publication How the shape vocabulary maps onto external standards, how it behaves in holder-selected disclosure protocols, and the shape in which the policy record itself is published.

Expression Alignment

Mapping between local shape vocabulary and external policy-expression and disclosure standards, including what does not map.

Policy Expression Crosswalk

The mapping between this model's selection, treatment, scope and grain concepts and external policy-expression standards, recording mapping strength, unmapped concepts and conformance claims explicitly withheld.

Which external policy-language terms correspond to this model's selection, treatment and grain concepts? interoperability
Which local concepts have no external equivalent and therefore require a profile? composition
Which conformance claims are explicitly not made, and what evidence would be needed to make them? validation
How is the crosswalk kept current when a mapped standard is revised? process

Selective Disclosure Protocol Fit

How a governed shape behaves in protocols where the holder rather than the source chooses what is revealed, and where accountability for the final shape then sits.

Which elements does the issuer make selectively disclosable, and which are always present? requirement
Who is accountable for the final disclosed shape when the holder makes the selection? ownership
How does a protocol's per-claim digest mechanism relate to this model's template fingerprint? identity
Which parts of a governed shape cannot be expressed in a holder-selected protocol at all? exception

Publication and Catalogue

The shape in which the policy record itself is disclosed to different readers.

Self-Applied Policy Projection

The projection under which the policy record itself leaves, since a disclosure policy is itself sensitive: a browsable summary, a template-only view for a bound recipient, and a full view for a reviewer.

Which projection of the policy record is published for unauthenticated browsing? access
Which parts of a policy record are withheld because publishing them would assist an attacker? security
How does a prospective recipient assess a shape without receiving any instance data? process
Which projection does a reviewer, auditor or supervisory authority receive, and on what basis? authority

Change and Fitness Shape-specific consequences of change: whether a change widens or narrows disclosure, what must be emitted so a past disclosure stays explainable, and which signals mean a shape no longer protects adequately.

Shape Change Semantics

Classification of a change between two policy versions and the reproducibility obligations that follow.

Disclosure Change Classification

Whether a change between two policy versions narrows, widens or leaves disclosure unchanged, which dependents it affects, and what each class requires before taking effect.

Is this change narrowing, widening or orthogonal with respect to what leaves? classification
Which compiled templates, bindings and recipients are affected by the change? relationship
What must be satisfied before a widening change may take effect, and who decides that? authority
How is a change treated when the policy text is unchanged but the treatment matrix or path grammar moved beneath it? provenance

Served Output Reproducibility

The minimum set of shape-identifying values that must accompany a served output so a past disclosure can be reconstructed and explained, and how long this model keeps them resolvable.

Which values must be emitted alongside a served output so its shape can be reconstructed later? requirement
Which model records those emitted values, and what is this model's obligation limited to? ownership
For how long must a superseded template stay resolvable, and what remains when it does not? retention
How are compilation time, effective time and disclosure time kept distinct in the emitted values? temporal

Fitness Signals

Detectable conditions that mean a previously acceptable shape may now under-protect, with hand-off to the models that act.

Shape Invalidation Signals

The observable conditions that make a shape unfit - upward reclassification, schema drift, changed adversary context, expired assurance, regime change - recorded as signals with safe interim behaviour and an explicit hand-off, not as a state machine.

Which observable signals indicate that this shape may now under-protect? event
What is the safe interim behaviour between detecting a signal and resolving it? state
Which model executes the resulting withdrawal or supersession, and what is handed to it? lifecycle
How is a long absence of signals distinguished from a shape nobody is monitoring? quality

Classifiers Filled

Family: World Models
Category: Cross-cutting context
Entry kind: mixin
Navigation path: NAV.XCT.DSC
Domain: XCT.DSC
Industry: Cross-industry
Tags: projectiondisclosurepolicyxct.dsc
Also called: S3

What it is Filled

A mix-in that specializes a generic governed policy (WM-KNW-012) with output-shape semantics only. It owns the declarative output shape (element selection, per-element treatment, record scope, aggregation grain), the order relation and least-disclosure combination over shapes, the compiled output template and its fingerprint, the binding of a shape to a target and audience as references, the declared identifiability and assurance properties of the shape, shape-specific change classification, and alignment with external policy-expression standards. It does not own policy identity, authority, approval, versioning state or conflict resolution; nor ownership, authorization, consent, classification, source-schema semantics, threshold or budget computation, audit records, or any runtime evaluation, decision, enforcement or execution over instance data. It holds no instance data.

In scope

Declarative output-shape definition: selection sets with a default treatment, per-element transformation technique and parameters, record-set scope, aggregation grain
Shape algebra: the subsumption test deciding whether one shape is no wider than another, and the meet operation combining two applicable shapes into a least-disclosure shape
Compiled output templates and their content fingerprints as reproducible derivations from a pinned policy version and pinned source schema version
Binding of a shape to a target, audience and effective window, expressed purely as references
Declared identifiability roles, assurance-method claims and cross-release linkability properties of a shape
Shape-specific change classification (narrowing, widening, orthogonal) and invalidation signals
Alignment of the shape vocabulary with external policy-expression and selective-disclosure standards
The projection under which the policy record itself is published

Out of scope

Generic policy identity, authority, approval, versioning state and conflict resolution (WM-KNW-012)
Ownership, stewardship and delegation over the underlying data (WM-XCT-001)
Authorization, purpose, parties and consent lifecycle (WM-XCT-002)
Audit records, append-only access trails and their ordering, immutability and retention (WM-XCT-004)
Cohort-floor and threshold computation, privacy budget accounting and aggregation assurance (WM-XCT-005)
Sensitivity classification schemes, criteria and class assignment (WM-XCT-020)
Source schema semantics, validation contracts and data quality (WM-DAT-004)
Runtime evaluation, decision outcome, explanation, obligation discharge and enforcement for a concrete request (WM-XCT-038)
Execution of transformations on instance data, key custody, and storage or transport of disclosed outputs
Legal determination of lawfulness: regimes are cited as constraints, never interpreted here

Why it exists Filled

Distinguishing features Filled

Defines the shape of what may leave a system: which elements, at what grain, with which treatments.
Holds no instance data; it operates on schemas and policies only.
Differs from the access contract (WM-XCT-002), which says who may read, and from runtime enforcement (WM-XCT-038).
References cohort floors and privacy budgets from their instruments instead of restating them.

What robots and AI may and may not do Filled

Must not

Serve data from an uncompiled or stale template.
Let unnamed elements leave by default.
Restate or lower a cohort minimum or privacy budget locally.
Store or cache source records.
Mark a shape as assured without a declared assurance method.

Only with a human decision

Approving a new or widened disclosure shape.
Accepting residual re-identification risk.
Classifying a disclosure change as material.

May

Compile an output template from a shape and a source schema.
Preview a shape against a schema to show what would leave.
Derive the least-disclosure shape for a stated purpose.
Detect source schema drift that makes a template stale.

Moral aspects Filled

Poorly designed shapes allow re-identification of people from aggregated or pseudonymised data.
Overly strict shapes can block legitimate research and public interest uses.
Small groups can be singled out; cohort floors protect minorities in data.

Who is affected

Data subjects
Data recipients and researchers
Data owners

Owners Filled

Steward

Name the policy register that issues authoritative master identifiers for projection policies, templates and matrices, and name the WM-KNW-012 authority under which shapes are approved; this model never mints a policy identity of its own.

Roles

Shape Author: Draft selection sets, treatments, record scope and grain with an explicit default and closure mode; Declare the path grammar, technique vocabulary and classification scheme versions the shape depends on; Record the identifiability role of every selected element and the assumptions behind it
Disclosure Reviewer: Test a candidate shape against the treatment matrix and the declared assurance method; Examine residual disclosure channels and record which are accepted and on whose authority; Return a review result; approval authority itself is exercised in WM-KNW-012 and is not held by this role
Template Compiler: Derive templates deterministically from pinned inputs and compute fingerprints; Refuse compilation when any pinned input is unresolvable or any path fails to resolve; Record its own implementation identity and version with every artifact produced
Schema Liaison: Monitor bound source schema versions for drift and raise invalidation signals; Request classification of newly appearing elements from WM-XCT-020 without assigning a class; Maintain the alias list that keeps prior paths resolvable after a rename
Catalogue Publisher: Apply the model's own publication shape to policy records and publish catalogue cards; Withhold parts whose publication would assist an attacker and record the reason; Keep published cards consistent with the current policy version or withdraw them
Model Steward: Maintain the expression alignment crosswalk and re-verify it when a cited standard is revised; Keep boundary notes current when a related model's scope moves, relocating any concept that drifts inward; Record conflicts, gaps and regional assumptions rather than resolving them silently

Links to other meta-models Filled

extends

WM-KNW-012 Governed Policy - Specialize the generic governed policy with output-shape semantics only. Policy identity, authoring authority, approval, version states, supersession and conflict resolution remain in the target; this model adds selection, treatment, record scope, grain, shape algebra, templates and shape-specific change classification.

references

WM-XCT-001 Ownership and Delegation - Carry a reference to the owner or delegated steward whose authority a shape presupposes, and to the party who accepted a named residual-disclosure risk. Ownership records, delegation instruments and their lifecycle stay in the target.
WM-XCT-002 Access Contract and Consent - Bind a shape to an access contract or consent grant by reference and record the audience class the binding serves. Authorization, purpose, parties and consent lifecycle stay in the target.
WM-XCT-004 Access Audit - Supply the provenance tuple - policy version identifier, template fingerprint, binding identifier, source schema version - that the target records with each served disclosure. Audit events, append-only trail semantics, ordering, immutability and retention stay in the target.
WM-XCT-005 Privacy Aggregation and Cohort Floors - Reference the applicable cohort floor, perturbation mechanism or composition budget for an aggregate or repeated-release shape. Threshold calculation, risk quantification, budget accounting and aggregation assurance stay in the target.
WM-XCT-020 Sensitivity Classification - Key the minimum-treatment matrix on classes assigned in the target and consume reclassification as an input signal. Classification schemes, criteria, membership and assignment stay in the target.
WM-DAT-004 Source Schema and Data Contract - Resolve path expressions and compile output templates against a pinned governed schema version, and report drift. Source semantics, structure and validation contracts stay in the target.
WM-XCT-038 Runtime Policy Enforcement - Provide the declarative shape and its applicability conditions as evaluation input. Selecting applicable policies, evaluating conditions, producing decisions and explanations, discharging obligations and applying a shape to instance data stay in the target.

aligned

W3C ODRL 2.2 Information Model and Vocabulary (external standard) - Map shape concepts onto Policy, Rule, Constraint, Party function and the anonymize, aggregate, derive and extract actions. Alignment only: ODRL actions carry no de-identification or field-selection parameters, so no conformance is claimed.
ISO/IEC 20889:2018 de-identification terminology (external standard) - Draw technique family names and the identifier, quasi-identifier and sensitive-attribute roles from a governed terminology instead of minting local terms. Conformance is not claimed because the normative text is paywalled and was not read clause by clause.
IETF RFC 9901 SD-JWT selective disclosure (external standard) - Align the encoding profile and residual-disclosure findings with a holder-selected disclosure protocol, including per-claim digests and decoy digests. Credential issuance, wallet holding and verification are separate subjects.

neighbor

WM-KNW-012 governed policy - The target owns policy identity, authoring authority, approval, version states, supersession and conflict resolution. This model adds only output-shape semantics and the shape-specific classification of a change; it never re-declares a policy lifecycle state machine.
WM-XCT-038 runtime enforcement - Using the XACML separation, this model is administration-side declarative content. Selecting applicable policies, evaluating conditions, producing a decision or explanation, discharging obligations and applying a shape to instance data all belong to the target.
WM-XCT-005 privacy aggregation - This model declares the grain, the measures and a reference to the applicable floor or budget instrument. Computing cohort floors, quantifying re-identification risk and accounting a differential-privacy budget belong to the target.
WM-XCT-020 sensitivity classification - The target owns the scheme, its criteria and the assignment of a class to an element. This model only keys a minimum-treatment constraint on an externally assigned class and reacts to reclassification as an input signal.
WM-DAT-004 source schema / data contract - The target owns source semantics, structure and validation. This model holds path expressions resolved against a pinned schema version, the resolution state of those paths, and the derived output template.
WM-XCT-004 access audit - This model defines only which shape-identifying values must be emitted with a served output. Writing, ordering, protecting and retaining the audit entry belong to the target; no audit-trail semantics are modelled here.
WM-XCT-002 access contract / consent - The target owns whether data may flow, for which purpose, to which party, under which consent. This model says only what the flow looks like and carries the binding reference.
WM-XCT-001 ownership and delegation - The target owns who holds an object and who may act for them. This model records only the reference to the authority required for a shape to be publishable and the party who accepted a residual-disclosure risk.
Serialization and interface layers (JSON, YAML, Markdown, Git, MCP, MongoDB) - Encodings are projections of one semantic shape. The model records which encodings are approved and whether semantics are preserved, and never treats a format as the definition of the shape.

What else AI and robots need to interact with it Filled

Sources Filled

ODRL Information Model 2.2 - World Wide Web Consortium (W3C)
ODRL Vocabulary & Expression 2.2 - World Wide Web Consortium (W3C)
NIST SP 800-188, De-Identifying Government Datasets: Techniques and Governance - National Institute of Standards and Technology (NIST)
NIST SP 800-226, Guidelines for Evaluating Differential Privacy Guarantees - National Institute of Standards and Technology (NIST)
45 CFR 164.514 - Other requirements relating to uses and disclosures of protected health information - U.S. Government Publishing Office / U.S. Department of Health and Human Services
RFC 9535, JSONPath: Query Expressions for JSON - Internet Engineering Task Force (IETF)
RFC 9901, Selective Disclosure for JSON Web Tokens (SD-JWT) - Internet Engineering Task Force (IETF)
ISO/IEC 20889:2018 Privacy enhancing data de-identification terminology and classification of techniques - International Organization for Standardization / International Electrotechnical Commission
ISO/IEC 27559:2022 Privacy enhancing data de-identification framework - International Organization for Standardization / International Electrotechnical Commission
Data Privacy Vocabulary (DPV) version 2.1 - W3C Data Privacy Vocabularies and Controls Community Group
eXtensible Access Control Markup Language (XACML) Version 3.0 - OASIS
Guidelines 4/2019 on Article 25 Data Protection by Design and by Default, Version 2.0 - European Data Protection Board
RFC 3339, Date and Time on the Internet: Timestamps - Internet Engineering Task Force (IETF)
JSON Schema: A Media Type for Describing JSON Documents (2020-12 dialect) - JSON Schema (IETF Internet-Draft draft-bhutton-json-schema-01)
45 CFR 164.514 Other requirements relating to uses and disclosures of protected health information - U.S. HHS (eCFR)
Regulation (EU) 2016/679 (GDPR) Article 5 — Principles relating to processing of personal data - European Parliament and Council / EUR-Lex
eXtensible Access Control Markup Language (XACML) Version 3.0 Plus Errata 01 - OASIS
XACML v3.0 Multiple Decision Profile Version 1.0 - OASIS
OData Version 4.01. Part 1: Protocol — §11.2.5 Specifying Properties to Return; §8.2.8.6 omit-values; §10 Context URL - OASIS
GraphQL Specification — Selection Sets, Fields, Client-specified response - GraphQL Foundation / Joint Development Foundation
HL7 FHIR R5 Search — _summary, _elements, SUBSETTED - HL7 International
JSON:API 1.1 — Sparse Fieldsets - jsonapi.org

Open questions

Decide whether the client-requested fieldset intersection rule (clip, reject, never expand) is a declarative parameter of this mixin or belongs entirely to WM-XCT-038 runtime evaluation, and only then author a node for it.
Re-author reusable named identifier treatment packs (Safe Harbor 18, limited data set 16, local packs) as pack references with jurisdiction and expansion semantics, without restating regulatory text inside the model.
Model completeness marking of a served payload (SUBSETTED-style token) so an incomplete projection cannot be used to overwrite a complete record, and determine whether that marking is a treatment, an encoding property or a downstream integrity contract.
Model a self-describing context handle that lets a recipient interpret a projected payload without out-of-band rules, and relate it to the existing compiled template fingerprint.
Resolve whether synthetic microdata generation is a treatment on this model or a sibling generative-privacy model, and whether a differential-privacy query interface is a grain class here with parameters owned by WM-XCT-005.
Gather primary evidence for non-tree media technique semantics: free-text redaction, image and audio masking, geospatial generalisation, and binary or large-object projection, none of which currently have modelled treatment semantics.
Non-tabular and non-tree media are addressed only through a generic locator hook: free-text redaction, image and audio masking, and geospatial generalisation have no modelled technique semantics.
Aggregate function semantics beyond declaration are absent; sum, count and distinct-count are named but their disclosure properties are not differentiated.
Synthetic data is named as a technique family but its fidelity and utility evidence, and the question of whether a synthetic output is a projection at all, are not modelled.
Dynamic query interfaces are covered only through release-set linkability; adaptive query answering and query auditing are not modelled.
A normative machine-readable encoding of the class-to-treatment matrix is left to the adopting Dimension, so mechanical checking is specified but not standardised.
Watermarking, traitor tracing and post-disclosure attribution of a leaked output are not modelled anywhere in this structure.
Utility loss caused by a shape is not measured, so no trade-off between protection and usefulness can be evaluated from this model alone.
Binary/media-stream and large-object projection (OData stream properties, FHIR Binary) has no first-class treatment profile in the cited specs.
Homomorphic or encrypted-field pass-through as a treatment class is mentioned in ISO 20889 cryptographic tools via secondary ITU-T X.1148 material; primary ISO text was not fully readable (paywalled). Marked gap.
Synthetic microdata generation as a leaving grain vs a sibling generative-privacy model is unresolved.
Differential-privacy query interfaces (NIST sharing model) as a grain class need epsilon/delta parameters that belong to XCT-005; the split may be too sharp for interactive query APIs.
SCIM attribute returnability, GraphQL @skip/@include directive policies, and MongoDB/SQL view projections were not fetched as primary sources in this pass.
National regimes beyond US HIPAA and EU GDPR (e.g. LGPD, PIPL, Australian Privacy Act APPs, UK DPA 2018 ICO anonymisation guidance) were not grounded in primary text.
Re-identification attack libraries and Disclosure Review Board operating procedures (NIST SP 800-188) are process-governance of de-identification programmes, only partly overlapping this mixin.
Format-preserving encryption and tokenisation vault semantics (PCI-adjacent) lack a primary source in this research set.

Machine files

Card as JSON Specification AGENTS.md Source Card backlog vercy.card/1

Provenance

world-models research · reviewable-draft

Built from: models/wm-xct-003-projection-disclosure-policy/spec.yaml, ver-cy/world-models/card-supplements/wm-xct-003-projection-disclosure-policy.json