Anonymous Technology Research In Data Publishing

Data publishing is to display database data to the user directly, for enhancing theexchange and sharing of data. While in various data applications to publish the database datain a direct way will expose private information of the data owners, thus will causeunnecessary mental injury or property damage to individuals. Therefore, it’s very necessaryto put the relevant data into anonymous processing before data publishing. Anonymousprocessing is a safe and effective method for data privacy protection, it can keep a goodbalance between data validity and data privacy preserving. The main idea of anonymousprocessing is to transform original data in some way, so that the attacker can not easily figureout the value of a sensitive tuple by analyzing the transformed data, and thus the specificinstance of which the sensitive information belongs can not be identified, so the goal ofprivacy preserving for individual information is achieved.In order to solve the existing problem of information leakage in multidimensionalsensitive attribute, this paper proposed an anonymous grading model for multidimensionalsensitive attribute based on loss links (g, K)-anonymous model by combining the existingmultidimensional barrel packet technology, and the formal description and correspondingimplementation algorithm for the model was also proposed.The specific work this thesis made were:Firstly, as the multidimensional barrel packet technology has the defect for approximateguessing attacks, a (g, K)-anonymous model considering multidimensional sensitive attributevalue distribution was proposed in this paper to solve this problem. As the model regulatedthe distribution number of the same sensitivity attribute value in groups, the model caneffectively resist the approximately guessing attacks.Secondly, the existing privacy preserving models mainly use the generalization andhiding method; the method needs to predefine a generalization tree for identifier attribute ofeach dimension, and was prone to over-generalization. A concept of attribute value exchangewithin the group was introduced to anatomy decomposition publishing in this paper based onloss links idea. By randomly exchange each dimension attribute values within groups, themethod can effectively reduce information loss due to generalization and hiding process in current models, and can solve the problem of background knowledge attack and existenceattack in existing models.Finally, the rationality and validity of the proposed method was verified throughexperiments, and the quality and efficiency of data publishing were also verified, acomparison between the proposed method and the largest multidimensional barrels groupingmethod was also made for analysis. The results showed that the (g, K)-anonymous algorithmperformed well in terms of process time and information losses, and can greatly improve theeffect of privacy protection for multidimensional sensitive attribute data.