How Much Is Enough? Choosing ε for Differential Privacy

By Jaewoo Lee, Chris Clifton. 2011

281 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Sociology and Political Science, Privacy & Security

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Sociology and Political Science

In Privacy & Security

Citations in the corpus

Foundation Models and Fair Use (2023, background): “While there is flourishing research on the topic (Lee & Clifton, 2011), none has studied this with the goal of leveraging DP to mitigate copyright-related risks.”
Privacy in Pharmacogenetics: An End-to-End Case Study of Personalized Warfarin Dosing (2014, neutral): “Lee and Clifton [30] recognize the problem of setting ε and its relationship to the relative nature of differential privacy, and later [31] propose an alternative parametization of differential privacy in terms of the probability that an individual contributes to the resulting model.”
Differential Privacy: An Economic Method for Choosing Epsilon (2014, background): “The most detailed discussion ofε we are aware of is due to Lee and Clifton [25].”
Cryptϵ: Crypto-Assisted Differential Privacy on Untrusted Servers (2019, background): “The choice of value for εB should be guided by social prerogatives [7, 61, 69] and is currently outside the scope of Cryptε .”
Differentially Private SQL with Bounded User Contribution (2019, methodology): “To represent privacy, there are well-established techniques [19, 27, 28, 37] and perspectives [40] in the literature.”
Privacy for All: Ensuring Fair and Equitable Privacy Protections (2018, background): “Choosing an appropriate value for is non-trivial and does not correlate with any privacy standards (Lee and Clifton, 2011).”
On Memorization in Probabilistic Deep Generative Models (2021, background): “We expect that future work will focus on this important question, and suggest that inspiration may be taken from work on choosing ε in differential privacy [57, 58].”
Differential identifiability (2012, background): “The ǫ in ǫ-differential privacy does not correspond to such a probability; it has been shown that for a given value of ǫ, the probability of identification can vary depending on data values, or even on values of individuals not in the data set [13].”