3.3 Securing AI Workloads PROFESSIONAL-CLOUD-SECURITY-ENGINEER Practice Questions (Page 5)

Questions 21–25

1. 21foundation · easy

   In an IaaS-based ML training environment, which combination of controls is essential to protect the trained model artifacts stored in Cloud Storage?

   • ACloud Monitoring and alerts
   • BVPC firewall rules and network tags
   • CCloud CDN and load balancing
   • DIAM permissions and encryption at rest
   Reveal Solution Discussion Select an answer first
2. 22foundation · easy

   Which practice is most effective for detecting adversarial attacks that cause a model to make incorrect predictions after deployment?

   • AContinuous monitoring of model performance and prediction confidence
   • BRestricting access to the training data
   • CUsing a larger training dataset
   • DEncrypting the model file
   Reveal Solution Discussion Select an answer first
4. 23application · medium

   A government agency is building a model to predict traffic patterns. They have a dataset containing GPS coordinates of citizens' vehicles. The privacy office mandates that the data must be anonymized such that an individual cannot be re-identified from the training data. What technique should the data engineering team use?

   • AAdd random noise to the GPS coordinates using a Laplace mechanism.
   • BGeneralize the GPS coordinates to a larger geographic area, such as a city block or zip code, and remove any direct identifiers.
   • CEncrypt the GPS coordinates using AES-256 and store the encryption key in Cloud KMS.
   • DTokenize the GPS coordinates by replacing them with a random alphanumeric string.
   Reveal Solution Discussion Select an answer first
5. 24foundation · easy

   Which practice is most effective for detecting anomalous behavior in an AI workload?

   • AEncrypting the model artifacts
   • BRestricting access to the training data
   • CSetting up alerts based on metrics and logs
   • DUsing a larger VM size
   Reveal Solution Discussion Select an answer first
6. 25foundation · easy

   Which technique transforms individual data values in a dataset so that the data cannot be traced back to a specific person, while still allowing useful analysis?

   • AAnonymization
   • BEncryption
   • CData minimization
   • DAccess control
   Reveal Solution Discussion Select an answer first