Differential privacy

Lecture 22: Differential privacy

-for differential privacy, understand what information is being protected and what information is not being protected

What information is being protected

• feels safe their are not much difference whether or not I take part in the survey

-understand how to compute the global sensitivity G of counting queries

Global sensitivity of a query Q is the maximum difference in

answers that adding or removing any individual from the dataset

can cause (maximum effect of an individual)

• Intuitively, we want to consider the worst case scenario

• If asking multiple queries, global sensitivity is equal to the sum

of the differences

Global sensitivity (G)

• How much difference the presence or absence of an indivual could make to the result
• can be calculated by look at the datasets

-understand the role of the privacy loss budget k

Privacy budget (K)

• how private we want the result to be (how hard for the attacker to guess the true result)
• A is the probability that result is R with me, B is the probability that results R without me. Choses k to gearantee that A <= 2^k * B
  • k=0 No privacy loss, low utility
  • k=hgih: larger privacy loss, high utility
  • k=low: low privacy loss, lower utility

-understand the role of G and k in terms of how much noise is added to the true query result

how much to add

• depending on G/k, that is, large G and small k allows more noise to be added.
• but the average value of the noise is 0

-understand that noise is added to the real query answer and this noise-added result is what will be released to the user. Understand how this protects the privacy of an individual

-understand that the amount of noise added is dependent on the ratio G/k. Larger G allows for more noise to be added and smaller k allows for more noise added