On Significance of the Least Significant Bits For Differential Privacy

We describe a new type of vulnerability present in many implementations of differentially private mechanisms. In particular, all four publicly available general purpose systems for differentially private computations are susceptible to our attack.

The vulnerability is based on irregularities of floating-point implementations of the privacy-preserving Laplacian mechanism. Unlike its mathematical abstraction, the textbook sampling procedure results in a porous distribution over double-precision numbers that allows one to breach differential privacy with just a few queries into the mechanism.

We propose a mitigating strategy and prove that it satisfies differential privacy under some mild assumptions on available implementation of floating-point arithmetic.

lsbs.pdf
PDF file

Publisher  ACM

Details

TypeInproceedings
> Publications > On Significance of the Least Significant Bits For Differential Privacy