K-MSD-WGT: A Local Recoding Algorithm Capturing Attribute Preferences
Jinling Song, Guohua Liu, Liming Huang · 2008
We address how to capture the attribute preferences and maintain higher precision of the anonymized table. At first, we define a new data dependency named k-multiset dependency (K-MSD), and show that if a dataset satisfies K-MSD then it also satisfies k-anonymity constraint. Then, we present minimal distance generalization (MDG) to construct K-MSD between attributes. In addition, we propose a local recoding algorithm: K-MSD-WGT, which performs MDG to k-violation values on attribute level based on the assigned preference weights. K-MSD-WGT can maintain higher precision and improve the utility of the released table.