Research On Privacy Preserving Methods For Sensitive Attributes In Data Publishing

Big data has a major impact on global production, circulation, distribution and consumption activities and economic operation mechanism, social life and national governance,to promote the development of large data industry, the primary task is to open the shared data, but these data may be work secrets and personal privacy and other sensitive information in data publishing by the enterprises or organization.Therefore, these data should be processed before publishing, and how to ensure that published data have high availability and sensitive information is not compromised becomes crucial.Firstly, this paper describes the research background?significance and research status of privacy, and details the way of attacker get user information?implement technology of anonymous model and so on. Secondly, after a deep study of K-Anonymity model and L-Diversity model, for both models were unable to resist background attacks and similar attacks, a difsimt),,( ulti-level anonymous algorithm for categorical sensitive attributes is proposed. The algorithm not only considers the link between sensitive property values, but also introduces the concept of similar values and distinct values to interpret realistic significance, while uses a strategy to make multiple close performance characteristics and multiple distinct performance characteristics of sensitive attribute values exist in an equivalence class which meets on the basis of the overall distribution. Once again, in the field of multi-sensitive attribute privacy protection, this paper has a in-depth study multidimensional bucket packet technology(MSB) and its proposed three algorithms,for MSB technology will have a high data loss and data occult rate when sensitive attributes increase, for this problem, an improved efficiency of MBF algorithm named Maximum Selectivity Bucket First is proposed. This method vertically partitions the original data table into a quasi-identifier attribute table and a number of sensitivesub-tables which is divided by correlation between the sensitive attributes to reduce the number of dimensions of the sensitive attributes, and then use ID connection between the pluralities of sub-tables. The algorithm can significantly reduce the rate of anonymity, and better protect the user's sensitive information. Finally, the experimental results verify the validity of these two algorithms, and compared with the traditional classical algorithms, these both algorithms can be a higher level of protection of user sensitive information.