Professional Cloud ArchitectDemo

By Google · Question Mode

QuestionQ11

Designing for security and compliance

Refer to the KnightMotives Automotive case study. KnightMotives wants to personalize its dealer experience for customers and has chosen to train its own AI models for personalized recommendations. The company will begin collecting personally identifiable information (PII) from customers for use as training data for the models. KnightMotives wants to ensure maximum security and worldwide compliance. You must ensure that the data is encrypted both at rest and while AI model training occurs, without affecting model accuracy. What should you do?

  • A Store the training data in BigQuery using column-level encryption. Train the model using Confidential GKE Nodes.
  • B Store the training data in BigQuery using column-level encryption Train the model on VertexAI notebooks using customer-managed encryption keys.
  • C Process all data with Sensitive Data Protection’s de-identification service. Replace any PII with a random string before storing it. Train the model using Confidential GKE Nodes.
  • D Process all data with Sensitive Data Protection's de-identification service. Replace any PII with a random string before storing it. Train the model on VertexAI notebooks using customer-managed encryption keys.
Explanation

BigQuery column-level encryption protects PII at rest, while Confidential GKE Nodes use hardware-backed memory encryption to protect data in use during model training without requiring application changes that could affect model accuracy. Replacing PII with random strings can remove or distort useful personalization signals, and customer-managed encryption keys on Vertex AI Workbench protect resources at rest rather than data being processed in memory.

Learn more

Community Discussion

No comments yet. Be the first to start the discussion!