tkl-Score: A Misuseability Score for Deciding How To Share Data

Abstract

As more and more data is collected, individuals and organizations are beginning to share their collected data to gain valuable insights. In doing so, these data stakeholders must be aware of the kind of impact that releasing data will have. Therefore, the misuseability scores M-Score and L-Severity have been developed to provide a measure of the potential damage to individuals and organizations when sensitive information from a dataset is released. This thesis introduces tkl-Score and its derivative tkl-Score_{max} which augments M-Score and L-Severity measures by increasing record scores when records are more identifiable in a source table with l-Distinguishing Factor, and also increasing record scores when sensitive attributes are less granular in a source table with l-Distinguishing Factor and t-Distinguishing Factor. In contrast, M-Score and L-Severity account for only record identifiability in a source table with k-Distinguishing Factor. tkl-Score and tkl-Score_{max} are shown to better characterize the risk of releasing records compared to M-Score and L-Severity due to accounting for sensitive attribute granularity.