Objective 4.2 Secure a Pipeline (includes Maintenance) AIP-210 Practice Questions (Page 4)

Questions 16–20

1. 16foundation · easy

   What is the primary purpose of data masking in an ML pipeline?

   • ATo encrypt data so it cannot be read without a key
   • BTo delete data that is no longer needed
   • CTo compress data to reduce storage costs
   • DTo replace sensitive values with realistic but fictitious data
   Reveal Solution Discussion Select an answer first
2. 17foundation · easy

   Which practice is most directly aimed at identifying security vulnerabilities in the code that builds an ML model?

   • AModel performance monitoring
   • BContainer image scanning
   • CData validation checks
   • DStatic application security testing (SAST)
   Reveal Solution Discussion Select an answer first
4. 18application · medium

   A software company's ML CI/CD pipeline automatically builds a model, runs unit tests, and deploys to a staging environment. The security team wants to add a gate so that a model with known vulnerabilities in its dependencies is never deployed to production. They also want to ensure that the model artifact is not modified between staging and production. What should be added to the pipeline?

   • AAdd a step that copies the model to production and then runs the unit tests on the production environment.
   • BStore the model in a private container registry and use the same tag for staging and production.
   • CAdd a manual approval step in the pipeline and require the security team to review the dependency list manually.
   • DAdd a dependency vulnerability scan step that fails the pipeline if critical vulnerabilities are found, and sign the model artifact with a code-signing certificate after the staging tests pass.
   Reveal Solution Discussion Select an answer first
5. 19application · medium

   A multinational corporation runs an ML platform where data scientists use Azure Machine Learning studio, and a service principal is used by the CI/CD pipeline to deploy models. An internal audit requires that all human access to the platform be individually attributable and that the service principal's permissions be limited to only the deployment task. What should the platform administrator configure?

   • ADisable MFA for the service principal but require it for data scientists, and assign the service principal the Reader role on the workspace.
   • BUse access keys stored in the pipeline's variables to authenticate the service principal, and grant all data scientists Owner access to the workspace.
   • CRequire Azure AD Multi-Factor Authentication (MFA) for all data scientists, and assign the service principal a custom role with only the 'Microsoft.MachineLearningServices/workspaces/models/read' and '.../endpoints/write' permissions.
   • DCreate a shared Azure AD account for the data science team and grant the service principal Contributor access to the entire resource group.
   Reveal Solution Discussion Select an answer first
6. 20foundation · easy

   What is the primary purpose of signing a model artifact?

   • ATo compress the artifact for faster download
   • BTo encrypt the artifact so only authorized users can read it
   • CTo make the artifact compatible with different ML frameworks
   • DTo prove the artifact's origin and ensure it has not been modified
   Reveal Solution Discussion Select an answer first