Significance of Term Relationships on Anonymization

Abstract

Sharing data provides great benefit to the research community. But disclosing identifiable, sensitive information such as medical records can cause irreparable damage. A number of methods have been proposed to anon Mize sensitive information. With some approaches, term relationships in the data may help to re-identify the original data given the de-identified data. This papers studies the significance of correlation in data and then analyzes the effect on anonymization techniques including t-plausibility and k-manonymity. Finally, we show how to address correlation in thet-plausibility model.