k-Anonymity using Two Level Clustering

Abstract

Data publising is becoming popular because of its usage and application in many fields. But original data have some sensitive information of individual whose personal privacy can be violated if original data is published . There are some agreements and policies which have to be fulfilled before publishing data . The techniques or protcols which preserve the privacy and retain useful information to apply data mining is nown as it is known as privacy preserving data publishing.k -anonymity is a technique to preserve privacy of individual while publishing data which still have useful information to apply data mining. To achieve k -anonymity local recoding algorithms gives less information loss but their execution time is more compared to global recoding algorithms. Their execution time mostly depends on for each cluster how they find the most suitable cluster to merge it,its linear search takes unnecessary time which can be reduce by find some most suitable cluster without linear search which we applied in our purposed algorithm. In our work, we used clustering at two levels , cluster at outer level contains inner clusters which are most likely to be merged. so to satisfy k value ,inner clusters merge within same outer cluster if still it do not satisfy k -anonymity then they merge with inner clusters of some other outer cluster, which other outer cluster is most suitable can be find without linear search and most of its inner cluster which still unsatisfied k -anonymity can be find without linear search. It this way we have reduced the execution time of our algorithm which it lesser than other efficient local recoding algorithm KACA and TopDown -KACA and other metrics such as distortion and discernibility gives similar resulted value as other local recoding algorithms.