Statistically useful anonymized/de-identified data sets will leak information when analytically combined with sufficiently rich and diverse exogenous data sets.

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.