MLA-C01 Question 158
Select 3You are designing a machine learning workflow using AWS services to train a model on customer purchase data. The dataset includes Personally Identifiable Information (PII) such as names, addresses, and phone numbers. Which of the following actions should you take to ensure compliance with data privacy best practices before training your ML model?
- A
Use AWS Glue to identify and redact PII data from the dataset.
- B
Encrypt the dataset at rest using AWS Key Management Service (KMS) before storing it in Amazon S3.
- C
Configure Amazon SageMaker to automatically detect and remove PII during model training.
- D
Leverage Amazon Macie to classify and discover PII in the dataset.
- E
Apply feature engineering techniques to directly encode PII fields into the model.
Show answer and explanation
Correct answers: A, B, D
Explanation
Ensuring compliance with data privacy best practices in AWS involves identifying, protecting, and securely processing sensitive information like PII. AWS Glue and Amazon Macie are effective tools for discovering and redacting PII, while encrypting data at rest using AWS KMS ensures its security. These steps should be performed before training the machine learning model. Encoding PII directly into the model or relying on SageMaker for automatic PII removal is not compliant with best practices.
- A. Correct.
Correct. AWS Glue provides capabilities for data preparation, including identifying and redacting PII data, which is a crucial step in ensuring privacy compliance before training the model.
- B. Correct.
Correct. Encrypting the dataset at rest using AWS KMS helps protect sensitive data from unauthorized access, which aligns with privacy and security best practices.
- C. Incorrect.
Incorrect. Amazon SageMaker does not have a built-in feature to automatically detect and remove PII during model training. PII removal should be handled during data preparation.
- D. Correct.
Correct. Amazon Macie is an AWS service designed to help discover and classify sensitive data, including PII, which can guide your data preparation efforts.
- E. Incorrect.
Incorrect. Encoding PII fields directly into the model without proper anonymization or redaction violates data privacy principles and could lead to compliance issues.