Research On Privacy Protection Of Dynamic Data Publication

As the emergence and development of data applications such as data publishing and data mining.How to protect people's privacy and prevent the disclosure of sensitive information becomes a serious problem.Anonymization techique is widely referenced in the protection of privacy,it can protect the provacy of our users and ensure data avaliability.So anonymization techniques become a research foucs in the area of privacy protection.In a static set of data annonymously published study,traditional K-Anonymity model has security problems and information loss problem.So a new data publishing algorithm based on cluster was proposed that on the basis of P-Sensitive-K model.In the algorithm,first to divide the sensitive attribute into different groups by sensitivity and set each group with different restriction in order to limit the same sensitive group's occurrence frequency in equivalence class and then divide data set into groups by cluster thought.Do all things for solving the security problems include background knowledge attack,homogeneity attack,probability attack and so forth.In a dynamic set of data annonymously published study,the M-Distinct is a quite excellent models that supports full dynamic set of data anonymization.Analysiss and study on M-Distinct model in this paper.First of all,as M-Distinct model no a certain limit at QI-Group-sensitive,so sensitive values in the QI-Group has a certain randomness.Resulting in M-Distinct model was prone to property attacks and the probability of attack.To this end,(M,CUS)-Distinct model was proposed.(M,CUS)-Distinct models require additional records in anonymity in the process,its QI-Group sensitive attribute value set must belong to the same set of CUS.Then dynamic data publishing,any records of FSUG increased viable paths in the figure.So as to reduce the chances of property attacks and the probability of attack.Secondly,because of the M-Distinct model in the distribution stage scoring record policy to distribute records,high time will cost.The proposed(M,CUS)-Distinct model creation phase generates compe disjoint barrel queues to ensure that each record can be stored in the record distribution phase in only barrel.This will reduce the time complexity of the algoritm.Finally,the experiment that based on real-world data sets showed,(M,CUS)-Distinct model in datat security,faked records,execution time of these three evaluation indicators are superior to the MDistinct model.