ρ-Kazanım: Mahremiyet Korumalı Fayda Temelli Veri Yayınlama Modeli
Özet
Data privacy is a difficult problem that tries to find the best balance between the privacy
risks of data owners and the utility of data sharing to the third parties. Anonymization is the
most commonly applied technique to overcome data privacy problems. The equivalence
classes, the natural outcome of anonymization process, are classified according to the data
utility in two main categories: Utility and Outlier Equivalence Classes (UEC, OEC). The
utility equivalence class contains records that have been suppressed by anonymization
techniques for privacy concerns. Meanwhile, the outlier equivalence class contains records
that have been fully suppressed by anonymization techniques resulting in no data utility. In
this study, ρ-Gain model, which focus on the effect of outlier equivalence class for increasing
data utility, was proposed. In the proposed model, k-Anonymity, ℓ-Diversity and t-Closeness
privacy models were used together with ρ-iterations to reduce the privacy risks. The Average
Equivalence Class metric was used to measure data utility. According to the findings
obtained from the study, the ρ-Gain model improved the data utility, but did not cause a
significant negative impact on privacy risk estimates. With the use of the proposed ρ-Gain
model as an anonymization technique, we have shown that the data utility has improved
while keeping the data privacy risk with no significant change.
