Skip to content

Designing a secure Delta Sharing strategy

Delta Sharing lets you securely share live Unity Catalog data (tables, views, volumes, notebooks, AI models) with users inside or outside your organization without copying data or requiring recipients to use Databricks. A secure sharing strategy relies on Unity Catalog's Shares and Recipients objects, combined with access controls, credential management, and activity monitoring to govern who can see what and how they authenticate.

1 · Learn the must-know

  • A share is a Unity Catalog securable object containing one or more tables, views, volumes, or other assets from a catalog/schema that you grant to one or more recipients.
  • Recipients can be Databricks-to-Databricks (authenticated via Unity Catalog identity, supporting Azure AD/Entra ID federation and no token management) or open sharing via a bearer-token-based credential file for non-Databricks recipients, which is less secure since tokens can be shared/leaked.
  • For open sharing, admins should set token expiration, rotate/rotate-and-revoke recipient credentials regularly, and use IP access lists to restrict where the credential file can be used from.
  • Use CREATE SHARE, GRANT SELECT ON SHARE, and CREATE RECIPIENT with fine-grained object-level grants so recipients only see the specific tables/views/columns/rows intended, leveraging views with row/column filtering for finer control before sharing.
  • All Delta Sharing grants and activity are governed by Unity Catalog and logged via unity catalog audit logs (system.access.audit), allowing providers to monitor recipient query activity and revoke access at any time by dropping the recipient or share grant.
  • Sharing across metastores/regions or with external (non-Databricks) parties should default to Databricks-to-Databricks sharing where possible for stronger identity-based security, reserving open sharing only when the recipient has no Databricks account.

2 · Check your understanding

Check this objectiveFree · always available

A provider at a healthcare company uses open (non-Databricks) Delta Sharing to send a dataset to a research partner that does not have an Azure Databricks or Unity Catalog account. The provider is concerned that if the partner's downloaded credential file is copied or leaked, it could be used indefinitely to pull data. What should the provider do when creating the recipient to reduce this risk?

Your objective map0 tried · 0 answered correctly · 77 untouched

What you have tried across DP-750's objectives, not a readiness score.

Set up and configure an Azure Databricks environment15-20% of the exam0 of 13 tried
Secure and govern Unity Catalog objects15-20% of the exam0 of 12 tried
Prepare and process data30-35% of the exam0 of 28 tried
Deploy and maintain data pipelines and workloads30-35% of the exam0 of 24 tried

3 · Keep going