Re:'Anonymised' data can never be totally anonymous, says study
by jandrewrogers on Sunday July 28, 2019 @03:28PM
(#20549370)
Statistically useful anonymized/de-identified data sets will leak information when analytically combined with sufficiently rich and diverse exogenous data sets. This is more of a "yet another example" than a new result. Techniques such as k-anonymity, chaffing, adding noise/randomness, differential privacy, etc significantly increase the computational cost and data requirements but not intractably so. The proliferation of vast sensing/event data sources provide nearly bottomless sources of exogenous data suitable for the purpose.
Anonymity is more polite fiction than rigorous fact these days.
Anonymity is more polite fiction than rigorous fact these days.