Researching For K-anonymity Relational Model And Related Theoretical Issues

k-anonymity privacy protection model which changes the data into uncertainty has become a trusted privacy protection model for the publisher of data. However, constrained by the privacy protection requirement, even if the optimal k-anonymity algorithm can not produce completely accurate data, k-anonymity privacy protection model has led to the reduction of the data using.The recipient reject the data as junk data. To make the data recipient treat and manage data in k-anonymity privacy protection model expediently and increase data availability, the k-anonymous data is studied from another point of view this article, namely, k-anonymity privacy protection model is taken as a kind of uncertain data, then deal with it using uncertain ways.Uncertain relation model, which extends every tuple of possible attribute values from the certain value in classic relation model to a value set, bases on the attribute values extension principle. This extension to uncertain relation model brings new requirements for data integrity constraints. Data dependencies are the formal representation for the mutual constraints condition between the values of attributes, functional dependencies are some of important influence data dependencies for the schema design and query in the database. In the uncertain relation, there are several possible values for every attribute in a tuple, so the mutual constraints for the values of attributes is more complicated than in the certain relation. Now, there are several following models that describe the uncertainty relations:1.tuple-ors,2.attribut-ors,3.C-tables,4.pNF2,5.Sprop,6,Sattr,7.Stuple,8.S2. Currently, there are only function dependencies and inference rules that based on tuple-ors model.But, the tuple-ors model does not have the schema definition, the function dependencies from the tuples-ors model can not guide the uncertain relation to standardized design. For the functional dependencies, due to the model which represents the uncertainty relation, function dependencies are in different forms and properties. In this paper, we presented two types of local functional dependencies, on the basis of the uncertain relation model-pNF2and then we proved inference rules. Take advantage of local functional dependencies proposed by this paper that can check out whether the uncertain relation violates the normal form design when the uncertain relation develops a possible world. It is the guidance of the uncertain relation model to normal form design as well as avoiding data redundancy and updating operation exception.At last, for data dependencies in the k-anonymity data, first, we defined a novel dependency, namely extended functional dependencies (EFDS), which extends the value of the determined properties get from the certain set, instead of the equality in traditional function dependencies, and is also different from the constraint of multivalued dependencies that the determined of properties in the multivalued dependencies is not related to the properties that not in the multivalued dependencies. Extended functional dependencies covers the semantics of a broad class of dependencies in databases, including functional dependencies, vertical functional dependencies, horizontal functional dependencies, and metric functional dependencies. It also describes the constraints of the k-anonymity uncertain data from the perspective of data integrity and instructs the selection of quasi-identifier in k-anonymity model. On the basis of extended functional dependencies, increase constraint conditions that the number of tuples of k-anonymity relational tables which quasi-identifiers are equal must be greater than k, and propose a k-extended functional dependencies, and discusses how to use the k-extended functional dependencies to check the operator of update in k-anonymous relational tables。...