How to Export Databricks Audit Logs for Security Review

Summary
A tutorial shows how to export Databricks audit logs with Spark SQL, save them as JSON in a Unity Catalog volume, and download them with the Databricks CLI. It also highlights activity to investigate and troubleshooting tips.
Key points
- Reading system.access.audit requires the system schema to be enabled and a SELECT grant; exporting to a Unity Catalog volume requires write access.
- A notebook can query audit events by date and action, write newline-delimited JSON to a volume, and download the files with the Databricks CLI.
- For security review, the article suggests checking unusual logins, token creation, permission or group-membership changes, off-hours configuration changes, and bulk downloads.
- The SQL Statement Execution API is an alternative, but its rows and separate column-name manifest require an extra conversion step.
- For workspaces without Unity Catalog, the article says to use audit log delivery to storage and DBFS paths instead.
Article Details
- Defense Focus
- Export Databricks audit and security-table logs for investigation, then review audit activity for unusual access, token, permission, configuration, or bulk-download behavior.
- Detection Methods
- Review logins and token logins from unusual IP addresses.
- Look for newly created personal access tokens.
- Review changes to permissions and group membership.
- Investigate workspace or cluster configuration changes outside normal hours.
- Review bulk downloads of notebooks or query results.
- Data Sources
- Databricks system.access.audit audit table
- Security data stored in Delta tables or other user tables
- Audit log files delivered to the workspace's own storage
- Platforms
- Databricks
- LogTotal
- Defensive Actions
- Enable the system schema and obtain SELECT access to read system.access.audit.
- Filter exports by date to limit the amount of data scanned.
- Export query results as newline-delimited JSON to a Unity Catalog volume, then download and combine only the part-*.json files.
- Sanitize the exported JSONL file and review it in LogTotal.
- For workspaces without Unity Catalog, use audit-log delivery to storage and DBFS paths.
Vendors
Products
DatabricksAs with Snowflake, “Databricks logs” can mean security data you keep in Delta tables, or Databricks’ own audit log in the system.access.audit system table.LogTotalBelow is how to get the data out, sanitize it, and investigate it for free in LogTotal.Unity CatalogQuery with Spark SQL, write the result as JSON to a Unity Catalog volume, then download it with the Databricks CLI.
