A Novel Dynamic KCi - Slice Publishing Prototype for Retaining Privacy and Utility of Multiple Sensitive Attributes

摘要

Data publishing plays a major role to establish a path between current world scenarios and next generation requirements and it is desirable to keep the individuals privacy on the released content without reducing the utility rate. Existing KC and KCi models concentrate on multiple categorical sensitive attributes. Both these models have their own merits and demerits. This paper proposes a new method named as novel KCi - slice model, to enhance the existing KCi approach with better utility levels and required privacy levels. The proposed model uses two rounds to publish the data. Anatomization approach is used to separate the sensitive attributes and quasi attributes. The first round uses a novel approach called as enhanced semantic l-diversity technique to bucketize the tuples and also determine the correlation of the sensitive attributes to build different sensitive tables. The second round generates multiple quasi tables by performing slicing operation on concatenated correlated quasi attributes. It concatenate the attributes of the quasi tables with the ID's of the buckets from the different sensitive tables and perform random permutations on the buckets of quasi tables. Proposed model publishes the data with more privacy and high utility levels when compared to the existing models.