A Study Of Efficiently Privacy Preserving Data Publishing Of Set-valued Data

Along with the development of computer technology and internet connectivity, many organizations collected and stored a large amount of data, it is important to release those data for study, however, these data often contains sensitive information about individuals. So, directly publishing these data will violate individual privacy, to solve this problem, privacy-preserving data publishing provides methods and tools for publishing useful information while preserving data privacy.As an important data type in privacy-preserving data publishing problem, set-valued data received some attention in research communities. Set-valued data is a kind of data in which a set of values are associated with an individual, such as market basket data. Unlike traditional relational data, set-valued data do not have distinctive quasi-identifiers and distinctive sensitive data, so traditional privacy principles and methods are not well-suited for set-valued data. We find that previous works on set-valued data may not protect individual privacy completely. So we develop a new privacy principle—(k,l)-anonymity, which can protect personal privacy completely. In addition to building a formal foundation for (k,l)-anonymity, we develop an algorithm which meets the privacy principle's requirements. Then, we show in an experimental evaluation that (k,l)-anonymity is practical and can be implemented efficiently.Like the static set-valued data, set-valued data stream is an import form of set-valued data, we proposed the problem of privacy-preserving publishing of set-valued data stream. To the best of our knowledge, this is the first paper about the problem. We firstly analyze the difference between privacy-preserving publishing of static set-valued data and privacy-preserving publishing of set-valued data stream, and describe the difficulty and importance of this problem. Then, we extend the (k,l)-anonymity privacy principle to the problem of privacy-preserving publishing of set-valued data stream. And then we developed an algorithm which meets the privacy principle's requirements to effectively handle the set-valued data stream. Finally, we analysis and confirm the algorithm's efficiency and effectiveness through an experiment in the actual data set.