Decoupled Data for Privacy Preserving Record Linkage with Error Management

Data from social networks are an excellent source of information for studying human behaviors and interactions. Typically, when analyzing such data, the default mode of access is de-identified data, which provides a level of privacy protection. However, due to its inability to link to other data, de-identified data has limitations with regard to answering broad and critically important questions about our complex society. In this paper, we (1) investigate the properties of information related to privacy, and (2) present a novel model of data access, decoupled data access, for studying personal data using these properties. Decoupling refers to separating out the identifying information from the sensitive data that needs protection. We assert that decoupled data access can provide flexible record linkage with error management while still providing the same level of privacy protection as de-identified data.