Semantic integrity and K-anonymity

The dataset in database have certain semantic commonly, and the semantic need to be satisfied with the form of some constrains, such as functional dependencies (FDs) and multivalued dependencies (MVDs). Nevertheless, the k-anonymity model may be destroyed the semantic integrity in the process of k-anonymization because of the incontinent generalizations. So, in this paper we address the issue of how to preserve the semantic integrity of dataset in the k-anonymization process. We define a new data dependency named k-multiset dependency (K-MSD), which can ensure a dataset satisfies k-anonymity constraint. In addition, we propose K-MSD algorithm to realize k-anonymization through constructing K-MSD between attributes, and propose K-MSD-AG algorithm to preserves FDs or MVDs as while as constructing K-MSD.