Research On Anonymity Technology For Microdata Publishing

Microdata publication offers many kinds of public data access services that cannot be easily provided with other data access techniques.Privacy is one serious issue within microdata publication. K-anonymization is the main technique used to implement privacy preservation during microdata publication.K-Anonymity model aims to granting the anonymized microdata privacy inference-proof ability to some extent.The quantitative exploration on this privacy preservation ability on K-anonymized data is indispensable for the wide application of this model in real public environment.This paper is in this topic research, the main research efforts are as followed:This paper introduces a multi-dimensional local recoding method based on the multi-dimensional generalization path and proposes two related K-anonymity algorithms,namely Filter K-anonymity algorithm and partial Filter K-anonymity algorithm.The two K-anonymity algorithms can prevent the potential privacy disclosure due to over-generalzation and other possible privacy violations.In comparison with classic Datafly and Incognito algorithm,the two algorithms offer more efficiency for both reducing anonymization cost and improving data precision.A formal multiple sensitive attributes microdata publication model is defined,named(k,s)-anonymity.It works beyond existed K-anonymity model and L-Diversity mechanism,which can effectively prevent multiple-attributed privacy violations.The corresponding(k, s)-anonymity implementation algorithm is proposed.This paper proposes a new anonymization cost metric(C)based on the anonymization coverage ratio.This metric can express some anonymization biases with the factor on each attribute in Q1,which is useable in special situations.At the end of this paper,some unsolved problems and future research directions are discussed.