Research On Date Publication Methids Based On Negative Representation Of Information

Most of the existing work on data publication focuses on how to conceal individuals’information during the publishing process and they are "positive" publication, which means directly publishing the processed version of the original data and the publishing data is essentially the real data. However, since the real data is published, attackers could retrieve individuals’information from the publishing data through certain types of attack, such as homogeneity attack. Then individuals’privacy is disclosed and the interest of the individuals is under great threat. The negative representation of information represents the information by its complementary set and it could be used to protect the privacy. Hence in this thesis the negative representation of information is introduced to data publication field and new data publication method (i.e. NPD) is proposed.Specially, the main work is this thesis includes the following aspects.(1) Since k-anonymity is a "positive" publication and it does not consider the sensitive information in the publishing process, the defects make k-anonymity vulnerable to some types of attack, such as homogeneity attack. Therefore, in this thesis the negative representation of information and k-anonymity are combined together and the (k, m)-anonNPD algorithm is proposed. The proposed algorithm publishes "negative" sensitive information and meanwhile generalizes the non-sensitive information. Through theoretical analyses and relevant experiments, the (k, m)-anonNPD algorithm is compared with k-anonymity on some benchmarks. It shows that the (k, m)-anonNPD algorithm enhances the privacy preserving ability of k-anonymity and it is practical at the same time.(2)L-diversity was proposed based on k-anonymity and it guarantees that the sensitive information in each equivalence in the publishing data has at least l distinct values. L-diversity overcomes the defect of k-anonymity, but it still publishes the real sensitive values. In this thesis the negative representation of information and l-diversity are combined together and the (l, m)-divNPD algorithm is proposed. The proposed algorithm publishes the negative sensitive information and could provide more diversity. Through theoretical analyses and relevant experiments, the (l, m)-divNPD algorithm is compared with l-diversity on some benchmarks. It shows that the (l, m)-divNPD algorithm enhances the privacy preserving ability of l-diversity and it is practical at the same time.(3) In this thesis two algorithms, namely (k, m)-anonNPD and (l, m)-divNPD, are proposed and they both have improved the effect of corresponding traditional data publication models (i.e.k-anonymity and l-diversity). However, the two algorithms have their own characteristics which make them different. Hence through relevant experiments the two algorithms are compared to analyze their advantages and disadvantages.In this thesis the negative representation of information is introduced to the data publication field and two algorithms, i.e. the (k, m)-anonNPD algorithm based on k-anonymity and the (l, m)-divNPD algorithm based on l-diversity, are proposed. The privacy preserving ability and utility of both algorithms are investigated through theoretical analyses and relevant experiments.