The impact of differential privacy on minority population data

DegreeMaster
StatusClosed
Supervisor(s)Univ.-Prof. Dr. Rainer Böhme

Description

Differential privacy (DP) has become popular as a data anonymisation technique that is increasingly adopted by official statistical institutes. However, differential privacy can lead to decreased data accuracy and well as biases, especially when post-processing is being used. Several researchers found that minority populations are disproportionally affected by this.

In this thesis, the student examines and compares existing work on this effect and tries to replicate it on census data from Europe using different DP mechanisms and post-processing operations.

Prerequisites

Privacy in statistical databases, interest in census data analysis

References

  • Dwork, C. A Firm Foundation for Private Data Analysis. Communications of the ACM, 54, 1 (2011), 86–95.
  • Christ, M., Radway, S., and Bellovin, S.M. Differential Privacy and Swapping: Examining De-Identification’s Impact on Minority Representation and Privacy Preservation in the U.S. Census. In IEEE Symposium on Security and Privacy. 2022, pp. 457–472.