Research On Privacy Preserving Data Publishing Based On Anonymity Models

In recent years,the rapid development of emerging technologies such as mobile Internet,cloud computing,Internet of Things,and smart cities has promoted the continuous growth of the total amount of digital information.The immeasurable value and information generated by the statistical research and application of relevant big data have greatly improved knowledge-based decision making of medical,commercial,energy,public transportation and other systems.Driven by this advantages,more and more data is shared and released among all parties.However,data is often directly or indirectly associated with personal information.If the data is released and used without any restriction and processing,personal information may be leaked,thereby infringing on their privacy.Therefore,it is very important to adopt solutions that can effectively maintain data availability while protecting personal privacy during the data release process.Firstly,this thesis studies the K-anonymity privacy protection of released data through clustering methods,and proposes a weighted K-member clustering algorithm.A series of weight indicators are designed to facilitate eliminating the influence of outliers before clustering,enhancing the correlation between records in the cluster,and reducing the information loss of published data caused by the K-anonymity clustering process.In order to speed up the grouping process of the clustering algorithm,the distance value is stored in the matrix,thereby reducing unnecessary calculations as well as the time complexity of the algorithm.Experimental results show that the proposed weighted K-member clustering algorithm outperforms other existing K-anonymity clustering algorithms in terms of clustering effect,data availability and operating efficiency.Secondly,in view of the privacy leakage problem that may be caused by the synonymous attack of sensitive attributes,this thesis proposes two effective K-anonymity publishing methods.The concept of synonymous linking of sensitive attributes is defined,and combine it with the distance between records and the information entropy to construct a new clustering evaluation index.Based on the above method,a micro-aggregation data publishing method is proposed to prevent synonymous attacks.A new type of l-diversity mechanism is designed to partition the data table and achieve privacy preserving data publishing.When creating equivalence classes,the characteristics of synonymous linkage of sensitive attributes are considered,and an l-diversity anatomy publishing method that can resist synonymous attacks is proposed.The dynamic update program is introduced into the two proposed algorithms to realize the insertion,deletion and modification of data.Experimental results show that the two algorithms proposed in this thesis maintain a better privacy protection effect while realizing the dynamic update of published data.