Objective 4.2 Secure a Pipeline (includes Maintenance) AIP-210 Practice Questions (Page 3)

Questions 11–15

1. 11foundation · easy

   In an ML pipeline, which component is most directly responsible for storing trained model files and their metadata?

   • AOrchestration engine
   • BFeature store
   • CModel registry
   • DData lake
   Reveal Solution Discussion Select an answer first
2. 12application · medium

   A healthcare startup packages its trained model as a Docker container and stores the image in Azure Container Registry (ACR) before deploying to Azure Kubernetes Service. The security team is concerned about a compromised base image and wants to ensure that only the exact container that was tested and signed is deployed to production. Which combination of practices should the ML engineering team implement?

   • AStore the image in a private ACR with a firewall rule, and manually verify the image tag before every deployment.
   • BEnable ACR's admin account and use its credentials in the AKS cluster to pull the image, then delete the image after deployment.
   • CUse a Dockerfile that pins the base image to a specific digest, and rebuild the image on the production cluster.
   • DScan the image with Microsoft Defender for Containers during CI, sign the image with a trusted key, and configure AKS to only pull images that have a valid signature.
   Reveal Solution Discussion Select an answer first
4. 13application · medium

   A government agency operates an ML pipeline that processes citizen data. The agency's security operations center (SOC) needs to be alerted when an unauthorized user attempts to access the model registry or when a pipeline run fails due to a permission error. Which logging and monitoring configuration should the ML platform team implement?

   • AStore all logs in the default Azure Storage account and have the SOC manually review them at the end of each month.
   • BEnable Azure Activity Log only and configure alerts for any 'write' operation on the workspace.
   • CEnable diagnostic settings on Azure Machine Learning to stream logs to a Log Analytics workspace, and create Azure Monitor alerts for 'UnauthorizedAccess' and 'RunFailed' events.
   • DUse a third-party SIEM to collect logs, but do not configure any alerts because the SIEM will automatically detect all threats.
   Reveal Solution Discussion Select an answer first
5. 14foundation · easy

   Why is regular patching of pipeline components important for security?

   • ATo make the pipeline run faster
   • BTo reduce the amount of data stored in the pipeline
   • CTo fix known vulnerabilities in software dependencies and libraries
   • DTo increase the accuracy of the ML model
   Reveal Solution Discussion Select an answer first
6. 15foundation · easy

   What is the primary purpose of audit trails in an ML pipeline?

   • ATo replace the need for encryption
   • BTo automatically improve model performance
   • CTo reduce the storage cost of the pipeline
   • DTo provide evidence of compliance with regulations and policies
   Reveal Solution Discussion Select an answer first