Improving the Utility of Differential Privacy via Univariate Microaggregation

摘要: Differential privacy is a model for anonymization that offers more robust guarantees than previous models, such as k-anonymity and its extensions. However, it often disregarded the utility of differentially private outputs quite limited, either because amount noise needs to be added obtain them or only preserved restricted type queries. On contrary, k-anonymity-like general purpose data releases make no assumption on uses protected data. This paper proposes mechanism offer with specific focus preservation Our proposal relies univariate microaggregation reduce needed satisfy differential privacy. The theoretical benefits are illustrated in practical setting.