‹ All AI-300 questions

AI-300 Practice Question 12: Managed Network Outbound Rule

Microsoft Machine Learning Operations Engineer Associate (AI-300) · Test 2 · question 2 of 10

Design and implement an MLOps infrastructurehard

A regulated team is configuring the managed virtual network of an Azure Machine Learning workspace. Their stated priority is minimizing the risk of data exfiltration, but training jobs still have to install two packages from the public pypi.org index, and the security team wants to be told up front about any billing consequence of the design. What should you configure?

Select one answer
Show answer and explanation

Answer:

  • A. Set the managed network isolation mode to allow_only_approved_outbound and add an FQDN outbound rule for the package host, accepting that FQDN rules add Azure Firewall charges

Microsoft's table of isolation modes lists allow only approved outbound for exactly this motivation: "You want to minimize the risk of data exfiltration". In that mode you open specific destinations with outbound rules, and the docs are explicit that "adding an outbound for a service tag or FQDN is only valid when the managed VNet is configured to allow_only_approved_outbound". A public host name such as pypi.org is expressed as an FQDN rule, and the cost warning the security team asked for is documented: "FQDN outbound rules are implemented using Azure Firewall. If you use outbound FQDN rules, charges for Azure Firewall are added to your billing." (option A).

- B is wrong because allow_internet_outbound lets the managed network reach the internet generally, which is the exfiltration path the team wants closed. Adding a private endpoint for storage hardens one dependency but leaves the open outbound path untouched. - C is wrong on the rule type. Service tags name ranges of Azure services, not arbitrary public hostnames, so a service tag cannot express pypi.org, and the claim that it avoids a firewall misstates where the cost comes from, since it is the FQDN rule type that pulls in Azure Firewall. - D is wrong because the managed virtual network is created and operated by Azure Machine Learning. Managed compute automatically joins it, and you shape its traffic through isolation modes and outbound rules, not by attaching your own NSG to subnets you don't own.

Reference: Managed virtual network isolation - Azure Machine Learning

Practice Test 2 ›

Study guides for AI-300