Note
Access to this page requires authorization. You can try signing in or changing directories.
Access to this page requires authorization. You can try changing directories.
Workspace outbound access protection helps safeguard your data by controlling outbound connections from items in your workspace to external resources. When this feature is enabled, Data Science items such as Machine Learning Experiments and Machine Learning Models can be created and used within the workspace.
Previously, Machine Learning Experiments and Machine Learning Models couldn't be created in workspaces with outbound access protection enabled. With this preview, these item types are now supported in protected workspaces.
Understanding outbound access protection with Data Science
Machine Learning Experiments and Machine Learning Models in Microsoft Fabric operate within the scope of a single workspace. Cross-workspace logging isn't supported for these item types. Because these items don't make outbound network connections on their own, no additional outbound access checks are required when outbound access protection is enabled.
The notebook code that generates Machine Learning Experiments or Models might access external data sources. Outbound access for notebooks is governed by the Data Engineering outbound access protection configuration, which controls how notebooks connect to resources outside the workspace.
Configuring outbound access protection for Data Science
To configure outbound access protection, follow the steps in Enable workspace outbound access protection. No additional configuration is required for Data Science items. After outbound access protection is enabled, Machine Learning Experiments and Machine Learning Models work within the workspace without further setup.
Exception mechanisms such as managed private endpoints or data connection rules aren't applicable to Data Science items, because these items don't initiate outbound connections.
Supported Data Science item types
These Data Science item types are supported with outbound access protection:
- Machine Learning Experiments
- Machine Learning Models
The following sections explain how outbound access protection affects these items in your workspace.
Machine Learning Experiments
With outbound access protection enabled, you can create and manage Machine Learning Experiments in the protected workspace. Experiments track runs, metrics, and parameters from notebook executions. Because experiment logging operates within the same workspace, outbound access protection doesn't restrict this functionality.
Machine Learning Models
With outbound access protection enabled, you can create and manage Machine Learning Models in the protected workspace. Models store trained model artifacts and version information. Because model creation and versioning operate within the same workspace, outbound access protection doesn't restrict this functionality.
Considerations and limitations
- Cross-workspace logging for Machine Learning Experiments and Machine Learning Models isn't supported in Fabric. All experiment and model operations must occur within the same workspace.
- Outbound access for notebook code that generates experiments or models is governed by the Data Engineering outbound access protection configuration. Make sure your notebook data sources are configured correctly if your workspace has outbound access protection enabled.
- For other limitations, refer to Workspace outbound access protection overview.