Federated Accountability Graph
The Federated Accountability Graph establishes a shared, cross-instance intelligence layer that aggregates relationship metadata across independent ClearTrace deployments into a unified, global view of systemic risk and accountability. Each participating instance contributes verified, public-safe metadata, enabling the network to map entities, relationships, and patterns of behavior across jurisdictions, sectors, and domains, without exposing sensitive data or compromising local data sovereignty.
This transforms ClearTrace from a set of independent systems into a coordinated, networked intelligence infrastructure, capable of identifying repeat actors, systemic patterns, and interdependencies that are not visible within isolated datasets.
What it is
Multiple independent ClearTrace instances opt-in to share graph metadata (not raw docs) to a global map service. The service aggregates nodes/edges and renders a live, explorable “systemic injustices” network.
Data shared (safe-by-default)
Nodes: entity_id_public, name, kind (org/person/project/case), country, sector, risk_grade, public_dossier_url. Edges: source_id_public, target_id_public, relation_type (subsidiary_of, funded_by, litigated_in, reported_by, partner_of), evidence_count_public, first_seen_at, last_verified_at. Flags: repeat_offender, under_investigation, verified_pct. No private evidence, PII, or unverified content is shared—only summaries + URLs to public dossiers.
System Model
Federation Model
Independent ClearTrace instances opt into a metadata federation layer Only public-safe, verified, and non-sensitive metadata is shared control over participation local governance policies A central Graph Hub aggregates and indexes contributed data for exploration and analysis.
Data Model (Public-Safe Metadata)
Nodes
entity_id_public — hashed, instance-scoped identifier kind — entity type (organization, individual, project, case) country — ISO country code risk_grade — low / medium / high / unknown public_dossier_url — link to source instance created_at — optional timestamp Edges
Flags
Privacy & Data Boundaries
The federation layer enforces strict data minimization:
No raw evidence is shared No personally identifiable or sensitive testimony data is exposed Only verified or public dossier-linked metadata is included All records reference back to originating instances Optional controls:
Per-edge redaction for jurisdictional compliance Instance-level participation policies Selective exposure based on verification status API (Instance → Hub)
Export
POST /federation/export
Payload:
Authentication:
scoped to federation:write Map Access
GET /federation/map
Returns:
Instance Health
GET /federation/instance/:id/health
Returns:
Identity & Deduplication
Each instance generates:
entity_id_public = sha256(instance_namespace + local_entity_uuid) Optional identity claims:
Deduplication strategy:
exact match on structured claims fuzzy match on name + country unresolved matches → possible_duplicate edge Steward review confirms merges.
Governance & Participation
Opt-In Federation
Participation is voluntary and configurable per instance Instances must accept a Federation Charter Federation Charter Defines
takedown and appeals processes Instance Reputation
Instances are evaluated based on:
Reputation influences:
weighting of graph signals trust in contributed metadata Graph Hub
The Graph Hub functions as:
It does not:
override instance-level governance User Interface (Global Graph)
Core Features
filtering by domain, sector, geography, and risk temporal exploration (time slider) relational mapping across entities and cases identification of high-centrality nodes and repeat actors Provenance
Each node/edge includes:
contributing instance count direct links to source dossiers Exploration Modes
curated “story paths” (narrative walkthroughs) MVP Implementation (2–4 Sprints)
public claims (legal_name, registry_no, country) relationship visibility levels SQL view (federation_graph_v) outputs public-safe node/edge payload lightweight service (Node/Go) validation (schema, signatures) storage (Postgres or graph DB) unresolved → flagged edges filters, search, provenance display onboarding and setup guide Versioning & Integrity
schema_version enforced across instances exported_at timestamp required bundle_hash ensures payload integrity hub maintains snapshot history for: Risks & Mitigations
Conflicting claims → parallel nodes + steward review Jurisdictional constraints → redaction + local policies Data poisoning → instance allowlist + anomaly detection Over-merging → conservative dedup strategy Roadmap
v2
signed provenance attestations v3
system-level pattern detection