Objective 4.2 Secure a Pipeline (includes Maintenance) AIP-210 Practice Questions (Page 5)

Questions 21–25

1. 21foundation · easy

   Which of the following is an example of a governance policy that an ML pipeline might need to enforce?

   • AThe choice of programming language for model training
   • BThe number of GPUs used for training
   • CThe color scheme of the monitoring dashboard
   • DData retention limits specifying how long raw data can be stored
   Reveal Solution Discussion Select an answer first
2. 22application · medium

   A financial services company runs a batch ML pipeline that ingests customer transaction data from an on-premises database, stages it in Azure Blob Storage, trains a fraud-detection model on Azure Machine Learning, and registers the model. A security audit requires that data be protected both while stored and while moving between these components. The data science team also needs to be able to read the staged data for exploratory analysis, but only the training compute should be able to write to the final model registry. What is the most appropriate set of controls to implement?

   • AEnable TLS for all data transfers, apply Azure Storage Service Encryption for Blob Storage, and assign separate Azure RBAC roles for read-only data access and write access to the model registry.
   • BEncrypt the data with a customer-managed key in Azure Key Vault, then grant all data scientists Contributor access to the workspace so they can manage the model registry.
   • CUse a shared access signature (SAS) token with full permissions for all pipeline components and rely on the corporate firewall to protect data in transit.
   • DApply Azure Disk Encryption to the on-premises database and use a private endpoint for Blob Storage, but do not configure access controls on the model registry.
   Reveal Solution Discussion Select an answer first
4. 23foundation · easy

   What is the primary purpose of versioning pipeline components?

   • ATo speed up model training
   • BTo automatically encrypt all pipeline data
   • CTo eliminate the need for access controls
   • DTo enable rollback to a known-good state if a change causes issues
   Reveal Solution Discussion Select an answer first
5. 24application · medium

   A retail company's ML team uses GitHub Actions to build and deploy a model training pipeline. Recently, a malicious dependency was introduced into the training code, and the team wants to prevent this from happening again. They also want to ensure that code changes are reviewed by a second person before the pipeline runs. What should they add to their CI/CD workflow?

   • AStore all dependencies in a single requirements.txt file and pin the versions, but do not add any scanning or review steps.
   • BAdd a dependency scanning step (e.g., Dependabot or Snyk) to the workflow, and configure a branch protection rule that requires a pull request with at least one approval before merging.
   • CUse a service principal with Contributor access to the repository, and configure the workflow to auto-merge all pull requests.
   • DAdd a step that runs the training code on a schedule, and require all team members to use the same personal access token for authentication.
   Reveal Solution Discussion Select an answer first
6. 25foundation · easy

   What is the primary purpose of integrating vulnerability scanning into the CI/CD pipeline for ML models?

   • ATo detect known security issues in dependencies and container images before deployment
   • BTo measure the model's accuracy on a test dataset
   • CTo ensure the model meets business requirements
   • DTo automatically fix all coding errors in the model training script
   Reveal Solution Discussion Select an answer first