Lakeflow Connect adds managed connector for Box audit logs
Databricks released the Box Audit Logs managed connector in Beta for Lakeflow Connect, enabling direct ingestion of enterprise audit events into Unity Catalog.
What changed
Databricks has released the managed Box Audit Logs connector in Lakeflow Connect in Beta. This feature enables teams to ingest enterprise audit events directly from Box into governed tables in Databricks.
Use cases
Imagine your security and compliance teams need to centralize file access logs, user activities, and administrative changes from Box alongside other enterprise logs. With this connector, you can:
- Ingest Box enterprise audit events incrementally and automatically.
- Govern ingested security logs directly under Unity Catalog.
- Schedule and orchestrate ingestion runs through Databricks Workflows or Declarative Automation Bundles (DABs).
How to use
Setup is handled through Unity Catalog connections and Lakeflow ingestion pipelines. The workflow consists of:
- Configure Box: Set up Box Server Authentication using the OAuth Machine-to-Machine flow with Client Credentials Grant (CCG).
- Create a Unity Catalog connection: A workspace admin configures the connection in Catalog Explorer or via API, enabling pipeline creation.
- Create the ingestion pipeline: Admins or data engineers create the ingestion pipeline using the UI, API, or Declarative Automation Bundles linked to the existing connection.
Limitations and requirements
- Stage: Available in Beta. A workspace administrator must activate Lakeflow Connect for Box Audit Logs on the Previews page.
- Authentication: Supports only OAuth M2M (Box Server Authentication with CCG). Basic authentication and OAuth U2M are not supported.
- Schema evolution: Supports new and deleted columns. Column renames are processed as an addition of a new column and deletion of the previous column. Changes to data types are not supported.
- Unsupported features: API-based row filtering and native SCD Type 2 are not supported.
Using it in a pipeline
Raw audit logs are ingested incrementally into a Bronze Delta table managed by Unity Catalog. Downstream pipelines can clean and parse activity types in the Silver layer for anomaly detection and aggregate metrics in the Gold layer for audit reporting.
- lakeflow connect
- ingestion
- box
- unity catalog
Box Audit Logs managed ingestion connector (Beta)
Summary and explanation generated by AI from the official release note. When in doubt, the platform's official documentation prevails.
Related
- DatabricksPublic Preview
DataFrame caching arrives on Databricks serverless compute
Now in Public Preview: call df.cache() and df.persist() on Databricks serverless compute to materialize expensive results into snapshots and speed up session reuses.
serverless compute · pyspark - DatabricksBeta
Configure Serverless Job Environments Using pyproject.toml
Serverless notebook and Python script tasks can now use a pyproject.toml file to define their job environment and manage project dependencies.
serverless · jobs - DatabricksBeta
Share serverless compute among notebooks in Asset Bundles
Allow notebooks and files in the same Asset Bundle to share serverless compute and an environment defined in the pyproject.toml file.
asset bundles · serverless