Provisional

Controls marked provisional are operator-drafted. No community governance group has directed, reviewed or endorsed them. This catalogue is not an Indigenous data sovereignty framework or product, and must not be presented as one. Published openly to invite critique, review and Indigenous-led governance direction.

Sovereign Knowledge Graph · Australian ISM Discipline · CARE & FAIR Controls

A knowledge graph for information that must not leak.

BlakSmith is an open, self-hosted knowledge graph built for protected Indigenous land title, cultural heritage and sensitive enterprise information, where custodians control the keys and rules rather than the platform operator. Paired with our vendor-neutral Indigenous Data Governance (IDG) Control Catalogue, it turns CARE, FAIR and Maiam nayri Wingara principles into 68 measurable controls in the discipline of the Australian Information Security Manual (ISM).

68
Proposed IDG Controls across 6 Domains
8
CARE & FAIR Data Handling Lifecycle Stages
40
Controls Backed by Automated Evidence Checks
12
Verified Framework Crosswalks (ISM, E8, PSPF, NIST)

1 · What BlakSmith Is & Why It Is Used

Built to hold, connect and govern records that ordinary systems put at risk

Organisations holding native title claims, Indigenous Land Use Agreements (ILUAs), cultural heritage records, tenure histories or sensitive corporate data usually scatter them across shared drives, spreadsheets and SaaS tools where administrators can read anything and provenance disappears on the first export. BlakSmith brings those records into one versioned, auditable knowledge graph governed by explicit cultural protocols and measurable CARE & FAIR controls.

Model 01

Everything is a node

Claims, determinations, parcels, heritage sites, people, organisations, agreements, systems and documents live as typed, classified records.

Model 02

Every relationship is an edge

Custodian authority, consultation obligations, tenure overlaps, system dependencies and data lineage are explicit and deterministically queryable.

Model 03

Every action is an event

Reads, refusals, consent changes, exports and key actions append to a SHA-256 hash-chained log with full temporal (“as-of”) reconstruction.

Model 04

Everything is governed

Access, encryption, presentation, sharing and AI use are enforced centrally before data leaves storage, never left to convention.

Use Case 01 · Land, Title & Country

Protected Indigenous land title & cultural heritage

  • Lineage & consultation: Answer which determinations, ILUAs and tenure interests affect a parcel, who the parties are, and who must be consulted before a dealing proceeds.
  • Time-travel (“as-of”) reconstruction: Inspect exactly what an interest or agreement looked like at the date of a determination without keeping plaintext copies of encrypted material.
  • Custodian-controlled access & sharing: Keep restricted cultural knowledge hidden from unauthorised search, counts and maps, and release data only under custodian-signed agreements.
  • Community transparency & return: Give communities a plain-language inventory of what is held about them, who accessed it for what purpose, and repatriate records with protocols intact.
Use Case 02 · Enterprise Discovery & Risk

Discovering, classifying & governing sensitive holdings

  • Find sensitive data across your estate: Discover and classify records across Microsoft 365 (SharePoint, OneDrive, Teams, Exchange), Google Workspace, Slack, cloud storage, file shares and databases.
  • Australian-aware detection: Built-in PSPF protective marking and Australian PII detectors (TFN, Medicare, ABN, IHI, passports, addresses, health records) with raise-only classification ratchets.
  • Blast radius & AI/MCP oversight: Connect people, infrastructure, software and AI systems to answer what breaks if a service fails and which AI agents or MCP servers can reach restricted data.
  • Zero-credential push ingestion: Ingest signed JSON evidence envelopes from internal systems without granting BlakSmith inbound credentials into sensitive environments.

Administrators cannot override protocols

Custodian authority is evaluated before clearance. Tenant owners, system admins, API keys, AI agents and emergency break-glass claims are refused access to protocol-bound material.

Per-custodian-group keys & erasure

Sensitive properties are sealed with AES-256-GCM envelope encryption before reaching Postgres (including version snapshots). Communities can hold their own key and erase data cryptographically.

Sacred-site & cultural presentation controls

Geometries are generalised to community-chosen grid cells, coordinate-shaped properties are stripped from packages, deceased-person names are suppressed, and restricted records fail closed.

Consent never launders into derivatives or AI

Consent is collective, purpose-bound and revocable. Derived datasets carry no inherited consent, and protocol-bound or restricted records are excluded from embeddings and external AI models.

Verifiable audit chain & database isolation

PostgreSQL row-level security isolates every tenant at the database layer, while an append-only SHA-256 hash chain lets communities verify access logs offline against anchored digests.

Self-hosted, onshore & deterministic

Licensed under Apache-2.0 and built for onshore Australian deployment via API, web UI and MCP. AI explains deterministic graph relationships; it never invents them or decides cultural sensitivity.

How BlakSmith and the IDG Control Catalogue fit together: The IDG catalogue below is deliberately vendor-neutral so any organisation can assess its people, processes and systems against CARE and FAIR. BlakSmith serves as an open worked example of how technical controls can be evidenced automatically, while recording every remaining gap honestly. BlakSmith is a technical foundation, not a finished Indigenous data sovereignty product, and awaits Traditional Owner direction.

2 · The Core Governance Shift

Separating sensitivity from sovereign authority

Most enterprise systems answer only one question: how harmful would a leak be? They attach a classification label, lock the record behind an administrator role, and quietly hand authority over community knowledge to whoever operates the system. IDG separates two distinct questions and assigns authority differently.

Question 01 · Organisational Security

How much harm would disclosure cause?

Answered by an ordered sensitivity ladder (public through restricted) that drives encryption, network transport, vetting, logging and secure environments. Automated pipelines may raise a classification, never lower it.

Mechanism
Sensitivity Classification
Held By
The Organisation
Question 02 · Sovereign Authority

Who has the authority to decide?

Answered by cultural protocols, Local Contexts labels and nominated community custodian groups. No system role, tenant owner, administrator or emergency break-glass claim can override community authority over protocol-bound material.

Mechanism
Cultural Protocol & Custodian Group
Held By
The Community
Tier 1 · Supreme Authority

The community over data about it

Decides collection, use, sharing, retention and destruction through its own processes. May adopt, strengthen, vary or reject any control for its own data.

Tier 2 · Catalogue Authority

Indigenous-led advisory group

Decides what the catalogue says and whether controls move from provisional to community-endorsed via append-only decision records.

Tier 3 · Maintainer Only

The operator (Yuma IT / BlakSmith)

Drafts provisional proposals, maintains open verification tooling and records group decisions on written instruction. Holds no cultural authority.

3 · CARE & FAIR Data Handling Lifecycle

Explore all 8 stages from planning to destruction & assurance

Select any lifecycle stage below (use mouse, touch, or Left/Right arrow keys) to inspect its proposed controls, CARE and FAIR alignment, and operational KPIs. Every control ID links directly to its full ISM-style specification in the manual.

Lifecycle Stage 1 of 8

Stage 1: Plan, Govern & Authority

CARE A1 CARE A2 CARE A3 CARE C1 CARE E2 CARE R1 CARE R2 FAIR R1.3

Establish community authority above organisational oversight, embed Aboriginal and Torres Strait Islander participation in data governance bodies, define separation of duties, and capture community data priorities and capability support needs before collection begins.

Proposed Controls (7)

  • IDG-GOV-001 Automated checks
    Community authority over Indigenous data
  • IDG-GOV-004 Organisational
    Governing body oversight below community authority
  • IDG-GOV-005 Organisational
    Aboriginal and Torres Strait Islander people in data governance
  • IDG-PPL-001 Automated checks
    Defined roles and separation of duties
  • IDG-PPL-002 Automated checks
    Custodian appointments and revocations come from the community
  • IDG-PPL-012 Organisational
    Support for community data capability, on the community's terms
  • IDG-PRC-014 Organisational
    Community data priorities

How This Stage Is Measured

KPI-LC1 Datasets with nominated community authority

(Datasets with named community authority / Total Indigenous datasets) x 100%

ML1
Register exists
ML2
100% mapped before use
ML3
100% system-enforced
KPI-LC1b Community data priorities actioned & reported

(Documented priorities reviewed & reported on schedule / Total recorded priorities) x 100%

ML1
Priorities requested
ML2
>= 90% reviewed on cycle
ML3
100% tracked with community sign-off

Existence hiding over open indexing (FAIR F4)

Where a community protocol requires it, unauthorised people cannot learn a restricted record exists in search, counts, maps or error messages (IDG-TEC-003).

Cryptographic erasure over persistent metadata (FAIR A2)

Communities can purge data and destroy per-group encryption keys on withdrawal or expiry; only the signed destruction attestation remains (IDG-TEC-007, IDG-TEC-016).

Fresh community consent over generic licences (FAIR R1.1)

Consent never launders into derived outputs or AI models. Every secondary use and derived dataset requires fresh community approval (IDG-PRC-004, IDG-TEC-017).

4 · Built for Two Audiences

Clear expectations for communities and receiving organisations

The framework is written in plain Australian English so communities can hold data holders to account and consuming entities know exactly what must be in place before and after receiving Indigenous data.

Audience 01

For Communities & Custodians

Plain-language clarity over what is held, how sensitive it is, who can reach it, for what purpose, and how it is returned or destroyed.

  • Transparent inventory: Accessible, protocol-aware inventory of every dataset held about the community (IDG-ASR-003).
  • Verifiable access logs: Hash-chained access records with external digests communities can check independently (IDG-TEC-011).
  • Real revocation & return: Community-held keys, cryptographic erasure and repatriation with protocols intact (IDG-TEC-006, IDG-PRC-013).
  • Right to review & vary: Community variations take precedence over the catalogue, and self-assessments go to the community for review.
Inspect CARE Coverage Matrix
Audience 02

For Consuming Entities & Enterprise Teams

Predictable, auditable controls for government, universities, health services, research institutes and enterprises receiving Indigenous data.

  • Familiar ISM discipline: Stable control IDs, applicability ladders, Essential Eight-style maturity levels and zero invented crosswalk numbers.
  • Reuse existing assurance: Crosswalks to ASD ISM, Essential Eight, PSPF, NIST and APPs show exactly what carries over and what IDG adds.
  • Proportionate onboarding: Three-tier recipient checklist scaled to sensitivity, from internal records up to restricted cultural material.
  • Honest technical evidence: BlakSmith serves as an open worked example that records real code, runnable checks and unvarnished gaps.
View Consuming-Entity Controls

5 · How to Measure It

Strict maturity scoring, operational KPIs & three-layer assurance

Governance claims are only as credible as their measurement. Following the ASD Essential Eight Maturity Model, each domain is scored at the level of its weakest applicable control, backed by cryptographically signed technical evidence and independent community review. Self-assessment is never presented as certification.

IDG-ML0

Not in place

Control is absent or incomplete. One unmet applicable control holds the entire domain at ML0.

IDG-ML1

Written & applied

Documented policy, registers, agreements and baseline safeguards are active across scope.

IDG-ML2

Enforced & reviewed

Controls are technically or procedurally enforced, logged, and reviewed on an agreed cycle.

IDG-ML3

Built-in & visible

Automated checks verify every change and communities can directly inspect or verify records.

Layer 1 · Continuous Technical

Signed Evidence Bundles

Runs every mapped unit test and verify script without leaking record content, signs the canonical JSON bundle with Ed25519, and anchors its SHA-256 digest in the tenant audit chain.

pnpm idg:evidence --tenant <slug> --key <pem>
Layer 2 · Annual & Event-Driven

Attested Self-Assessment

Validates completeness, scores every domain by weakest applicable control, locks fixed non-certification attestation wording, and records the digest in the audit chain.

pnpm idg:assessment assessment.yaml --record --tenant <slug>
Layer 3 · Independent Oversight

Community Review & Health Checks

Communities review assessments independently (accepted, queried or rejected) with their own record reference, complemented by impartial partnership health checks.

IDG-ASR-001 · IDG-ASR-004 · IDG-ASR-005
Live control counts and BlakSmith automated verification coverage across all 6 domains
Domain Focus Proposed Controls Automated Checks Code Only Organisational / Manual
GOV Governance and authority 5 2 1 2
PPL People 12 9 0 3
PRC Process 16 10 0 6
TEC Technology 17 16 0 1
ASR Assurance 5 3 0 2
CON Consuming entity 13 0 0 13

6 · Verified Framework Crosswalks

Zero invented control numbers across 12 established standards

Every external control identifier is verified against official publications (and re-checked automatically against ASD and NIST OSCAL catalogues via pnpm idg:verify-sources). No mapping is marked equivalent, because every IDG control adds collective community obligations beyond organisational security.

CARE Principles

Global Indigenous Data Alliance (Sep 2019)

Primary people & purpose anchor (C, A, R, E) mapped on every control.

FAIR Guiding Principles

Wilkinson et al. 2016 / GO FAIR

Metadata, authorisation, provenance & community standards alongside CARE precedence.

Maiam nayri Wingara

Australian Indigenous Data Sovereignty Collective (2018)

Five Australian Indigenous data sovereignty principles mapped on every control.

ASD Information Security Manual

Australian Signals Directorate (Sep 2026, OSCAL v2026.09.4)

58 verified data & security controls reconciled via automated OSCAL checks.

ASD Essential Eight

Australian Signals Directorate (Nov 2023)

Inspiration for strict weakest-link ML0–ML3 maturity scoring plus MFA, admin & backup links.

PSPF Release 2026

Attorney-General's Department

Australian Government protective security, personnel vetting, sharing & disposal links.

NIAA AFGID v1.0

National Indigenous Australians Agency (May 2024)

APS Framework for Governance of Indigenous Data actions mapped as non-binding guidance.

Privacy Act 1988 (APPs & NDB)

OAIC

Individual privacy and breach obligations extended to collective community rights & harm.

NIST CSF 2.0 & SP 800-53 Rev. 5

NIST (OSCAL Verified)

62 verified security and privacy controls enabling enterprise assurance reuse.

TRUST Principles

Lin et al. 2020

Repository transparency, responsibility, user focus, sustainability and technology.

ISO/IEC 38505-1

ISO / IEC (Paraphrased, Provisional)

Governing body evaluate-direct-monitor oversight placed explicitly below community authority.

Local Contexts TK & BC Labels

Local Contexts Hub

Machine-readable cultural authority notices and protocols travelling with every record.

7 · Inspect, Test & Critique Openly

Explore the full branded manual and machine-readable artefacts

Every artefact below is generated deterministically from the validated YAML sources in docs/idg/. Use the branded HTML manual for executive briefings and A4 PDF printing, the CSV/Markdown catalogues for governance tooling, and the self-assessment template to benchmark current practice.

Open Full Branded ISM Manual