Spatial Analysis of Privacy Measured Through Individual Uniqueness Based on Simple U.S. Demographics Data

Abstract

Previous studies reveal that, using U.S. census data, over 60% population of the U.S. could be uniquely identied with a combination of gender, zip code, date of birth attributes in 1990 and 2000. This thesis extends these studies to examine spatial variation of individual uniqueness in 2010 at dierent scales and regions in the U.S. In this thesis, I use spatial and non-spatial statistics to study the spatial patterns on both global and local scales. Specically, I provide 1) the comparison of national level uniqueness between 2000 and 2010, 2) the investigation of spatial variation of uniqueness in different regions and at dierent scales, 3) the identication of local uniqueness clusters outliers and 4) the evaluation of urban-rural divides on individual uniqueness segregation. On the global scale, the comparison between 2000 and 2010 reveals that, although overall individual uniqueness changes little, the individual uniqueness of middle-age group members has signicantly decreased. The study of regional differences finds that low individual uniqueness for college-age population are spatially homogeneous despite that the overall uniqueness are spatially heterogeneous. The analysis at different scales discloses that overall uniqueness decreases, and the dierences between age-group uniqueness reduce, when geographical scales focus on the cores of urban area. On the local scale, the results indicate an urban-rural divides of individual uniqueness segregation. The Clusters and Outliers Analysis nd that places where low individual uniqueness cluster the most are also very urbanized area. The average individual uniqueness of urban area is computed as 58.02% whereas that of rural area is computed as 88.43%. This means, if a person is from an urban area, given the zip code, gender and date of birth information, he/she is much less likely to be identied uniquely. This study offers contributions to geographic information privacy, particularly relevant to reverse geocoding and related spatial aggregation techniques used in census data.