A Note on the Misinterpretation of the US Census Re-identification Attack

Abstract: In 2018, the US Census Bureau designed a new data reconstruction and re-identification attack and tested it against their 2010 data release. The specific attack executed by the Bureau allows an attacker to infer the race and ethnicity of respondents with average 75% precision for 85% of the respondents, assuming that the attacker knows the correct age, sex, and address of the respondents. They interpreted the attack as exceeding the Bureau's privacy standards, and so introduced stronger privacy protections for the 2020 Census in the form of the TopDown Algorithm (TDA). This paper demonstrates that race and ethnicity can be inferred from the TDA-protected census data with substantially better precision and recall, using less prior knowledge: only the respondents' address. Race and ethnicity can be inferred with average 75% precision for 98% of the respondents, and can be inferred with 100% precision for 11% of the respondents. The inference is done by simply assuming that the race/ethnicity of the respondent is that of the majority race/ethnicity for the respondent's census block. The conclusion to draw from this simple demonstration is NOT that the Bureau's data releases lack adequate privacy protections. Indeed it is the purpose of the data releases to allow this kind of inference. The problem, rather, is that the Bureau's criteria for measuring privacy is flawed and overly pessimistic.