Data Entrepreneurs' Synthetic PUF: A Working PUF as an Alternative to Traditional Synthetic and Non-synthetic PUFs
Joshua Borton, Elizabeth C. Hair · 2013
The nature of Medicare Claims data makes it infeasible to apply usual methods of creating synthetic or non-synthetic PUFs due to data complexity, numerous identifying variables (IVs), and the difficulty of computing disclosure risk under any assumed intruder IV knowledge as knowledge may increase over time due to growing public availability of personal information. In view of this, we consider a two-prong strategy: creating a working PUF with high confidentiality at the cost of analytic utility, coupled with DUA-controlled access to microdata for testing the applicability of procedures developed for the working PUF in final analysis. The working PUF -- termed as data entrepreneurs’ synthetic PUF (DE-SynPUF) -- has high pseudo-analytic utility in that it retains the original database structure and is thus useful to data entrepreneurs for application development and for researchers in training and some familiarization with the data on which the PUF is based. The DE-SynPUF was created by treating beneficiaries and individual claims, with no explicit preservation of intra-claim relationships. Moreover, all claims were subject to post hoc treatment in order to reduce risk when kanonymization is used for de-identification. An application of DE-SynPUF to 2008-10 beneficiaries and claims data is presented.