Privacy considerations in ML Python - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to import the library used for differential privacy in machine learning.

ML Python

from diffprivlib import [1]

Drag options to blanks, or click blank then click option'

Amodels

Bdatasets

Cprivacy

Dmechanisms

Attempts:

3 left

💡 Hint

Common Mistakes

Importing 'models' instead of 'mechanisms'.

Using 'privacy' which is not a valid module.

The mechanisms module provides tools to add differential privacy to machine learning models.

2fill in blank

medium

Complete the code to add Gaussian noise for privacy to a numeric value.

ML Python

noisy_value = original_value + [1](0, 1)

Drag options to blanks, or click blank then click option'

Arandom.randint

Brandom.uniform

Crandom.gauss

Drandom.normalvariate

Attempts:

3 left

💡 Hint

Common Mistakes

Using uniform noise instead of Gaussian.

Using randint which adds integer noise.

random.gauss adds Gaussian noise with mean 0 and standard deviation 1, useful for privacy.

3fill in blank

hard

Fix the error in the code to ensure the model does not memorize sensitive data by limiting training epochs.

ML Python

model.fit(X_train, y_train, epochs=[1])

Drag options to blanks, or click blank then click option'

A1

B1000

C50

D500

Attempts:

3 left

💡 Hint

Common Mistakes

Using too many epochs causing overfitting.

Using zero epochs which means no training.

Using fewer epochs like 1 helps prevent overfitting and memorizing sensitive data.

4fill in blank

hard

Fill in the blank to create a dictionary comprehension that filters out data points with sensitive attribute value 'Yes'.

ML Python

filtered_data = {k: v for k, v in data.items() if v[1] 'Yes'}

Drag options to blanks, or click blank then click option'

A>

B!=

C<

D==

Attempts:

3 left

💡 Hint

Common Mistakes

Using '==' which keeps only 'Yes' values.

Using '<' or '>' which are not meaningful for strings here.

We want to keep data where the value is not equal to 'Yes', so use != 'Yes'.

5fill in blank

hard

Fill all three blanks to create a dictionary comprehension that maps user IDs to their masked emails only if their consent is True.

ML Python

masked_emails = {user[1]: email[2] for user, (email, consent) in users.items() if consent [3] True}

Drag options to blanks, or click blank then click option'

A[::-1]

B.lower()

C==

D[:3] + '***'

Attempts:

3 left

💡 Hint

Common Mistakes

Using reverse slicing which doesn't mask properly.

Not filtering by consent.

Not converting email to lowercase.

The user ID is masked by showing first 3 chars plus stars, emails are converted to lowercase, and we filter only users with consent True.