Privacy Lect5

Question 1

What is the comparison for datasets in PII (personable identifiable information)?

Answer 1

1 anonym and 1 known

Question 2

What is the goal in linking external information for PII?

Answer 2

Re-identify entities

Question 3

How to prevent PII from being leaked?

Answer 3

Use encrypted email, HTTPS

Question 4

GDPR (General data protection regulation) requires?

Answer 4

Transparency in profiling customers

Question 5

Privacy contradictory to Blockchain

Answer 5

The right to be forgotten

Question 6

Data anonymization example

Answer 6

Netflix case with iMBD

Question 7

Machine learning example

Answer 7

based on what your bought, supermarket can guess if and when you’re pregnant

Question 8

How to protect data privacy in data sharing?

Answer 8

Privacy Enhancing Technology

Question 9

Types of PET (2)

Answer 9

K-Anonymity and Differential Privacy

Question 10

What is K-Anonymity?

Answer 10

Model for releases of microdata that seek to prevent record re-identification by hiding each original record within a group of k-indistinguishable anonymized records

Question 11

Dataset said to be k-anonymous if?

Answer 11

Combination of values appear for at least k different records.

Question 12

2 factos in k-anonymity

Answer 12

I-diversity and t-closeness

Question 13

How to make dataset k anonymous?

Answer 13

Use opensource tool ARX

Question 14

K-anonym doesn’t provide privacy if:

Answer 14

Sensitive values lack diversity, and attacker has background knowledge.

Question 15

What is differential privacy?

Answer 15

Limit knowledge that users obtain from query records

Question 16

How to do differential privacy?

Answer 16

Add noise

Question 17

Differential privacy protects against?

Answer 17

Differencing attacks, linkage attacks and reconstruction attacks

Question 18

Differential privacy doesn’t protect against?

Answer 18

Only protect private information and not general information

Question 19

What does Federated Learning in ML enable?

Answer 19

A model to be trained by multiple machines, where training and data reside in a local machine. Results are encrypted and shared with collaborating machiens.

Question 20

What is Synthetic Data in ML?

Answer 20

Data which is mimicked to include properties similar to real data collected from actual business transactions.

Question 21

What is transfer learning in Machine learning?

Answer 21

Allow pre-trained models to be used by others for developing a new model in a similar field.

Question 22

Usage of Zero-Knowledge Proof

Answer 22

E-voting, E-auction, privacy-preserving blockchain, nuclear disarmament