Anonymization

Updated: 2020-10-19

Differential Privacy

The industry standard for anonymizing data. E.g. it is used by the US Census Bureau for US census.

K-anonymity

Weaknesses:

  • vulnerable to attackers with knowledge of the dataset or who can fake data into the dataset
  • not robust when the same dataset is released multiple times or when the dataset is too homogeneous.
  • easy to be applied incorrectly, e.g. counting events instead of unique users