Enhanced Stratified Sampling: A Method for Privacy Preserving Big Data Publishing

Big data has proven to be extremely useful in the development of modern data analysis, despite the large storage and time-consuming nature of incorporating a set into one's study. However, as more data is used and published, privacy concerns can arise through means as simple as overlaying multiple of these massive datasets. In this paper we propose a method, Enhanced Stratified Sampling (ESS), that integrates the strengths of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$k$</tex>-anonymity, <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$l$</tex>-diversity, and t-closeness to offer a solution for privacy preserving big data publishing. ESS generates multiple subsets of the data and evaluates them based on privacy loss and information loss, selecting the optimal set for publication. This approach not only enhances privacy but also maintains high data utility, making it suitable for various applications that necessitate large-scale data analysis without sacrificing privacy.