Enhanced slicing models for preserving privacy in data publication

Abstract

Privacy preservation in publishing of microdata has been studied extensively in recent years. Microdata contain records each of which contains information about an individual entity, such as a person, a household, or an organization. Several anonymization techniques, such as generalization, bucketization and slicing have been designed for privacy preserving microdata publishing. That generalization loses considerable amount of information, especially for high dimensional data. Bucketization does not prevent membership disclosure and does not apply for data that do not have a clear separation between quasi-identifying attributes and sensitive attributes. Slicing have a drawback when more number of similar attribute value and the sensitive value may present in the different tuples may give the original tuple while performing the random permutation. The utility of the dataset is lost by generation the fake tuples. Thus enhanced slicing models have designed to overcome the drawbacks of slicing. The suppression slicing is done by suppressing any one of the attribute value in the tuples and then perform the slicing. Thus utility is maintained with minimum loss by suppressing only very few values and privacy is maintained by random permutation. The next model is Mondrian slicing in this the random permutation is done with all the buckets not within the single bucket. Thus same utility of the original dataset is maintained.