Obtaining (Ɛ,δ)-differential privacy guarantees when using the Poisson distribution to synthesize tabular data

Jackson, James and Mitra, Robin and Francis, Brian and Dove, Iain (2024) Obtaining (Ɛ,δ)-differential privacy guarantees when using the Poisson distribution to synthesize tabular data. In: Privacy in Statistical Databases – PSD2024 :. Lecture Notes in Computer Science . Springer, Cham, pp. 102-112. ISBN 9783031696503

Text (Poisson_and_Differential_Privacy-4)
Poisson_and_Differential_Privacy-4.pdf - Submitted Version
Available under License Other.
Download (487kB)

Text (Jackson_et_al_PSD_2024-accepted)
Jackson_et_al_PSD_2024-accepted.pdf - Accepted Version
Restricted to Repository staff only until 11 September 2025.
Available under License Other.
Download (579kB)

Abstract

We show that differential privacy type guarantees can be obtained when using a Poisson synthesis mechanism to protect counts in contingency tables. Specifically, we show how to obtain (ϵ, δ)-probabilistic differential privacy guarantees via the Poisson distribution’s cumulative distribution function). We demonstrate this Poisson synthesis mechanism empirically with the synthesis of the ESCrep data set, an administrativetype database that resembles the English School Census.