Prepare and test your skills
Prepare and test your skills
Free sample question, answered and explained below. The Professional Cloud Security Engineer practice bank has 1,343 more like it. Try 10 free: instant score, no account.
A financial enterprise deploys custom-trained classification models to Vertex AI to predict sensitive loan outcomes. A cloud security engineer is tasked with establishing security controls across the model deployment pipeline and inference serving layer to meet the following requirements:
Which combination of controls should the security engineer implement?
This option configures model serving inside Google Distributed Cloud databases (such as AlloyDB) using google_ml_integration SQL extension functions (google_ml.predict_row and google_ml.embedding), storing third-party provider credentials in Secret Manager.
google_ml_integration, custom transform functions are only supported for text-embedding models and cannot be passed when registering generic model endpoints.google_ml_integration extension is intended for database queries invoking remote AI endpoints, not managing Vertex AI prediction infrastructure.A database application running in Google Distributed Cloud needed to enrich SQL rows with text embeddings generated by textembedding-gecko using a database service account.
This architecture establishes end-to-end model governance, cryptographic provenance verification, API output defenses, and private network perimeter security using native Vertex AI capabilities.
This solution directly addresses data privacy vulnerabilities at the inference tier while enforcing supply-chain integrity controls and private network transport.
This approach suggests registering remote models from BigQuery ML into the Vertex AI Model Registry, applying generative AI Model Armor templates to suppress prediction outputs, and using standard public prediction endpoints protected by IAM.
The workload was deploying generative AI models (such as Gemini) requiring prompt/response safety filtering and floor policy enforcement against adversarial text injection.
This option focuses on running batch evaluations via Dataflow against ground truth datasets, using multi-region Cloud Storage Object Versioning for artifact history, and deploying default prediction containers without post-processing.
The primary goal was measuring model quality degradation (model drift) over time by comparing offline batch inference outputs against ground-truth datasets.
Answer them yourself and get your score the moment you finish. No account needed.
Full access includes
Keep the momentum going with these hand-picked practice scenarios
Last updated Validated