Masking is effective protecting individual’s privacy by obscuring direct identifiers. However, it leaves datasets susceptible to more sophisticated attacks, such as linkage attacks. In a linkage attack, quasi-identifiers are used in combination with external data to re-identify individuals. Generalization - also known as blurring - allows you to replace a data value with a less precise one via binning, reformatting, rounding or truncating.
When you ensure individuals are not unique across all quasi-identifiers in a dataset, you gain the upper hand defending against linkage attacks. K-anonymity ensures that the smallest number of indistinguishable individuals is a group of size ‘k’. That means every individual in a dataset is indistinguishable from at least k-1 others.
With Manual Generalization, you define the generalization rules. Privitar’s advanced capabilities allow you to set the k-anonymity threshold in accordance with your tolerance for linkage and re-identification risks for each specific data use. The Privitar Platform automatically drops rows that are in clusters that fall below thresholds so that you know k-anonymity is achieved.
With Automatic Generalization, you define k-anonymity cluster size and allow the Privitar Platform to dynamically determine the generalization rules to use. Privitar adjusts the blurring used to achieve k-anonymity for all quasi-identifiers – without the need to drop records. You maintain the greatest precision and include all of the data to achieve maximum data utility.