Research On Anonymized Privacy Preserving Publishing Of Data Streams

The privacy preservation of data publishing releases much more accurate data informa-tion while protecting the privacy. Anonymization is one of the most important implemen-tation methods. Currently, how to preserve the privacy in the streaming environments hasbecome an emerging hot spot.In this paper, we propose a balance-based privacy preserving algorithm of streamsnamed B-CASTLE, and extend the Anatomy to streaming environment, which is calledSAnatomy.Firstly, we study the identical metric mapping of the mixed attributes in streamingtuples, which is the foundation of clustering. In this paper, we present a semantic taxonomy-based distance metric between two categorical attributes. This metric measures the distancemore precisely, and improves the cohesion of clusters and utilization of data finally.Secondly, we work how to dynamically adjust and re-cluster the equivalence classes ortuples. In this paper, we do not consider the distribution of the stream but the time stamps ofthe tuples arrived, and then construct the dynamically adjusting and re-clustering frameworkof B-CASTLE, which limits the maximum number of tuples in one cluster, and improvesthe whole publishing process through merging the clusters partially.Thirdly, we study the preprocessing and publication. We introduce the generalizedpublishing algorithms of CASTLE at first, and improve and optimize the relevant pub-lishing strategy combined with B-CASTLE, the experiments results given in the end. Be-sides, we propose SAnatomy to preserve privacy of data streams, and experiments show thatSAnatomy has a better performance on data real-time processing and utilization.The researches and improvements on the algorithms of privacy preserving publishinghave certain practical significance, which promote further research of privacy protecting.Besides, we give a novel solution of publishing streams using Anatomy, which providesreference for other similar applications.