Recursive partitioning and summarization

Wahbeh Qardaji, Ninghui Li
2012 Proceedings of the 7th ACM Symposium on Information, Computer and Communications Security - ASIACCS '12  
In this paper we investigate Recursive Partitioning and Summarization (RPS), a practical framework for data publishing that satisfies differential privacy. While there have been several negative results concerning non-interactive differentially private data release, we show that such results do not necessarily mean that such release is impossible. To that end, we propose a data release framework that leverages current advances in differentially private query answering to synthesize an
more » ... dataset. We show that since each query only affects a sub linear number of tuples, we are able to guarantee differential privacy. To evaluate the efficacy and general applicability of our approach, we experimentally evaluate the utility of our framework in three domains and several real and synthetic datasets. All our results indicate the applicability of our framework to general data release.
doi:10.1145/2414456.2414477 dblp:conf/ccs/QardajiL12 fatcat:z2nui4m2lbdqngf2zfdno5rrpa