Question 340
A reviewer wants to validate the completed action “Authenticate training sources and apply integrity, provenance, and validation controls to imported datasets.”. Which observation is the strongest direct evidence for the original issue: A data ingestion model accepts unauthenticated training records, letting external parties poison downstream classifications.
Explanation
Correct Answer
For “A data ingestion model accepts unauthenticated training records, letting external parties poison downstream classifications.”, the requirement is for evidence, not another remediation proposal. “Unauthorized records are quarantined and model-quality checks detect manipulated input samples.” tests whether “Authenticate training sources and apply integrity, provenance, and validation controls to imported datasets.” actually achieved the intended effect.
Incorrect Answers
A positive result here relates to enforcing invoice-level authorization per authenticated API request rather than trusting numeric IDs; it leaves authenticating and validating training-data ingestion to prevent adversarial data poisoning untested.
This observation would verify “A new application stores reset tokens in plaintext logs visible to all support analysts.”, not the particular condition “A data ingestion model accepts unauthenticated training records, letting external parties poison downstream classifications.”. It cannot independently confirm that “Authenticate training sources and apply integrity, provenance, and validation controls to imported datasets.” produced “Unauthorized records are quarantined and model-quality checks detect manipulated input samples.”.
The measured result “Crafted serialized objects cannot instantiate arbitrary types or trigger unexpected server actions.” does not match the outcome of “Authenticate training sources and apply integrity, provenance, and validation controls to imported datasets.” for “A data ingestion model accepts unauthenticated training records, letting external parties poison downstream classifications.” and instead belongs to the alternate problem “A deserialization endpoint creates application objects from untrusted input without type restrictions.”.