Research Of Privacy Preserving Methods For Multiple Sensitive Attributes

With the rapid development of information technology and network, more andmore information is stored and published on the web, which makes informationsharing becomes more simple and convenient. Data publication as a means of sharingof resources, while facilitating data exchange and sharing of information, will result inthe disclosure of private information. Data release agencies usually take sometechnical means to remove personal identification or sensitive data, but this does notguarantee the security of personal privacy information, through a number of openconnections between data sources often results in unexpected operation of the privacyinformation leakage problem. With the requirements of increasing privacy, privacypreserving has become a hot research direction. At present, most of the anonymoustechnology deals with the data with a single sensitive attribute, but in the real world,data to be released often involves multiple sensitive attributes, so the research ofprivacy preserving methods for multiple sensitive attributes has great significance.Firstly, this paper studies the existing privacy preserving techniques, especiallythe privacy protection technology in data publishing, and makes a conclusion of thetwo major hot spots in this field: the anonymous model and the anonymoustechnology. Then the current models and techniques are studied. Via comparativeanalysis, the paper makes a summary of the advantages and disadvantages of themodels and techniques.Secondly, the paper researches on the privacy preserving techniques for the datawith multiple sensitive attributes, and points out the shortcomings of themulti-sensitive bucketization in the protection of privacy. For the multi-sensitivebucketization cannot resists the background knowledge attacks and the similarityattacks, a new anonymous model-(l1, l2,... ld)-uniqueness is released, and thecorresponding anonymization algorithm is given. The new model processes eachsensitive attribute independently, and breaks the one-to-one relationship between thesensitive properties, so it can withstand a certain amount of background knowledgeattacks. What’s more, the sensitive attribute values in a same group are required tohave different sensitivity level, and then the new model can withstand the similarityattacks. We use the actual data sets to do experiments on the new anonymous modeland anonymity algorithm, and the results show that the anonymity model has asmaller loss of information, and it can protect the privacy of individuals well, andimproves the security of the data releasing.Finally, this paper researches the distribution publem of the sensitive attributevalues in the publication of multiple sensitive attributes,and proposed the L-coverage grouping method based on clustering.Experimental results on the realworld datasets show that the new grouping method can ensure the security of datapublishing and keep more information available.