Permissions

Foundation implements a sophisticated and flexible permissions system that combines role-based access control (RBAC) with attribute-based refinements to create a highly customizable security model.

Permission Scopes

The platform uses a hierarchical permission model that operates across multiple organizational levels:

  • Foundation Level: The highest boundary, which refers to each individual platform instance.

  • Organization Level: Divisions within Foundations (departments, business units, data domains)

  • Mesh Level: Collaborative spaces where teams share and build data and use cases.

  • Entity Level: Each data product, data system, data source, ...

Roles

Users cannot be directly assigned policies; they must be assigned roles, which contain the policies. Currently, each role should be assigned only one policy.

The system defines several pre-configured roles, each tailored to specific responsibilities:

Administrative Roles:

  • IT Superadmin: Full platform control in order to set up Foundation and connect to data sources.

  • IT OrgUnit Admin: Organizational unit management in order to provide technical support on connection to other systems, user management and data pipelines.

  • Data Superadmin: Complete data governance and data management control

Data Governance Roles:

  • CDO: Full visibility across Foundation

  • Data Owner: Full control over data assets within an Organization

  • Data Steward: Data quality and governance enforcement within an Organization

  • Data Mesh Manager: Controls mesh governance and sharing policies

Operational Roles:

  • ML Model Developer/Manager/Viewer: AI/ML asset management with varying permissions

  • Data Product Developer/Editor/Viewer: Graduated access levels for data product management

  • Data Consumer: Can consume data through the APIs

  • Asset Owner: Can perform read, update and delete actions on the asset they own (data sources, data systems, applications, data products,...)

Policy-Driven Permissions

What makes Foundation's approach particularly powerful is how policies bridge roles and actual permissions. Rather than hard-coding what each role can do, the system uses policies as an intermediary layer:

  1. Roles are assigned policies rather than direct permissions

  2. Policies contain sets of permissions with defined scopes

  3. Permissions can be dynamically adjusted based on contextual attributes

For example, a Data Product Developer role might have different effective permissions based on:

  • Which organization they belong to

  • The classification level of the data (public, confidential, restricted)

  • The specific mesh they're operating within

  • Time-based constraints or geographical restrictions

Attribute-Based Refinements

The system extends beyond traditional RBAC by incorporating attribute-based access control (ABAC) capabilities. This means permissions can be refined based on:

  • User attributes: Department, location, clearance level, project assignment

  • Resource attributes: Data classification, ownership, creation date, business criticality

  • Environmental attributes: Time of access, network location, device compliance

  • Action attributes: Read vs. write, bulk operations, export capabilities

Permission Flows In Practice

Consider a typical data sharing scenario:

  1. A user from OrgUnit1 needs access to a data product in Mesh1

  2. Their base role (e.g., Data Product Viewer) is evaluated

  3. Policies associated with that role are checked

  4. Attributes are evaluated (user's org unit, data classification, mesh policies)

  5. The effective permission is calculated and granted/denied

The Mesh Manager can configure whether:

  • All data products in their mesh are automatically accessible to mesh members

  • Users need to request individual permissions for each data product

  • Certain data products require additional approval workflows

Governance Through Flexibility

This flexible model enables sophisticated governance scenarios:

  • Data sovereignty compliance: Policies can restrict data access based on geographical attributes

  • Temporal access: Permissions can be time-bound for contractors or project-based work

  • Dynamic classification: As data classification changes, permissions automatically adjust

  • Cross-mesh collaboration: When data products from Mesh1 and Mesh2 are shared in Mesh3, both Mesh Managers must approve, and their respective policies are preserved

Enterprise Integration

The system seamlessly integrates with enterprise identity providers through:

  • SSO Integration: Connect to providers like Azure AD/Entra ID

  • Group Mapping: Automatically map enterprise user groups to Foundation roles

  • Attribute Synchronization: Pull user attributes from corporate directories

This architecture ensures that Foundation can enforce fine-grained access control while remaining manageable at scale, adapting to complex organizational structures and regulatory requirements without sacrificing usability or performance.

Last updated