A company wants to develop an ML model by using tabular data from its customers. The data contains meaningful ordered features with sensitive information that should not be discarded. An ML engineer…

Question

Options

• AUse Amazon Made to categorize the sensitive data.
• BPrepare the data by using AWS Glue DataBrew.
• CRun an AWS Batch job to change the sensitive data to random values.
• DRun an Amazon EMR job to change the sensitive data to random values.

• A
  7% (2)
• B
  74% (20)
• C
  4% (1)
• D
  15% (4)

Explanation

AWS Glue DataBrew is purpose-built for data preparation and includes native data masking transformations (substitution, shuffling, hashing, encryption) that preserve the structure and order of tabular data - critical when features are meaningful and ordered. It provides a no-code/low-code interface specifically designed for this kind of pre-processing handoff between teams.

Why the distractors are wrong:

• A (Amazon Macie): Macie discovers and classifies sensitive data (like PII) in S3 - it identifies what's sensitive, but does not mask or transform it. It's a detection tool, not a transformation tool.
• C (AWS Batch with random values): Replacing sensitive values with random values destroys the statistical relationships in the data, which would corrupt a meaningful ordered feature. Proper masking (e.g., tokenization or format-preserving encryption) must preserve value utility.
• D (Amazon EMR with random values): Same fundamental flaw as C - randomization breaks the integrity of ordered, meaningful features. EMR is also general-purpose big data processing, not specialized for governed data masking workflows.

Memory tip: Think of the flow - Macie = Find it, DataBrew = Fix it. When a question says "mask before another team builds the model," that's a preparation step → DataBrew. If the question said "discover what's sensitive," that would be Macie.

No community discussion yet for this question.