Research On Privacy Preserving Data Publishing For Multi-sensitive Attribute Based On Clustering

Data release plays an important role in data management, data mining and information sharing applications. The rapid development of the Internet makes the collection, dissemination and analysis of large amounts of data more convenient. At the same time, but also to bring a threat to users’ privacy. In real life, there are many agencies need to regularly released data, for example, medical hospital regularly publish statistics, listed companies regularly publish financial statements, and so on. The presence of a large number of sensitive personal information in these data, if disclosure would cause immeasurable loss. The actual scene, there is more than one attribute sensitive data, so we study the existence of more than one privacy-sensitive property imminent release of the data, especially for multi-sensitive attribute data released multidimensional numerical sensitive attributes personalized privacy protection issues, and more it is to explore the hot spots, analysis of this issue often requires clustering of data processing, so called for the cluster.Firstly, the paper analyzes the existing data published in a variety of ways to protect privacy, including anonymous privacy protection model, anonymizing technologies, application clustering anonymous methods and so on. Draw advantages and disadvantages of different models and anonymous anonymous technologies.Secondly, the paper published the data privacy protection technology more sensitive properties in depth, pointing out the traditional multi-sensitive attribute data distribution method has limitations on the protection of privacy. Posted privacy protection method for multi-dimensional numerical sensitive attributes data, rarely considered sensitive attribute value attribute value weight individual privacy issues. It proposes personalized privacy protection method based on clustering and multidimensional weighted barrel packet. Firstly, the individual value of each property is divided by clustering dimensional numerical sensitive attributes to multiple cluster, and then construct a weighted multidimensional cube bucket sensitive property value, the data in the table is mapped to the corresponding record multidimensional bucket, weighted by considering the maximum dimensions of the priority capacity selection algorithm to select multidimensional data records in the bucket, build to satisfy packet l-diversity, and finally will get a packet identifier quasi generalization, the group released in the form of an anonymous list. Experimental results show that the release of anonymous data table, meet l-diversity at the same time, to avoid similar attacks, with lower information loss and lower hidden rate, and higher data owners to define important release of the records can be reached personalized privacy protection.Finally, the paper published the multi-sensitive attribute data in multidimensional attribute value sensitive data anonymous Posted personalized technology in depth, the traditional method, there is little value considering weight-sensitive property values of the weights heavy and sensitive issue. Proposed based on weighted clustering and personalization(data record weights) to select the degree of anonymity algorithm which utilizes a minimum degree of personalization data record selection priority of thought, select the data records constituted to satisfy packet l-diversity, and finally will be quasi-generalization packet identifier, the packet publish anonymous table form, to personalized privacy protection. Experimental results show that this method to ensure data privacy and personal data or privacy needs of the owner, but also has a high quality of data dissemination.