nerdexam
Microsoft

DP-100 · Question #433

You have a dataset that contains records of patients tested for diabetes. The dataset includes the patient's age. You plan to create an analysis that will report the mean age value from the…

The correct answer is C. 0.5. In differential privacy, epsilon must be a positive value, and a smaller positive epsilon adds more noise to the data, providing stronger privacy protection and minimizing the risk of exposing actual data.

Explore data and run experiments

Question

You have a dataset that contains records of patients tested for diabetes. The dataset includes the patient's age. You plan to create an analysis that will report the mean age value from the differentially private data derived from the dataset. You need to identify the epsilon value to use in the analysis that minimizes the risk of exposing the actual data. Which epsilon value should you use?

Options

  • A-1.5
  • B-0.5
  • C0.5
  • D1.5

How the community answered

(26 responses)
  • A
    4% (1)
  • C
    92% (24)
  • D
    4% (1)

Why each option

In differential privacy, epsilon must be a positive value, and a smaller positive epsilon adds more noise to the data, providing stronger privacy protection and minimizing the risk of exposing actual data.

A-1.5

-1.5 is a negative value, which is mathematically invalid for an epsilon parameter in differential privacy and would cause an error in any compliant implementation.

B-0.5

-0.5 is also a negative epsilon value, which is invalid; differential privacy requires epsilon to be greater than zero.

C0.5Correct

Epsilon is the privacy budget parameter in differential privacy and must be a strictly positive number. A value of 0.5 is smaller than 1.5, meaning more statistical noise is injected into the query result, which makes it harder to infer the true underlying data and thus minimizes re-identification risk. Choosing the smallest valid positive epsilon provides the strongest privacy guarantee.

D1.5

1.5 is a valid positive epsilon but provides weaker privacy than 0.5 because a larger epsilon permits less noise and brings the output closer to the true data values.

Concept tested: Epsilon selection in differential privacy analysis

Source: https://learn.microsoft.com/en-us/azure/machine-learning/concept-differential-privacy

Topics

#Differential Privacy#Epsilon#Data Privacy#Privacy-Preserving Analysis

Community Discussion

No community discussion yet for this question.

Full DP-100 Practice