Anonymously Publishing Univariate Time-Series: With focus on (k,P)-Anonymity
Erik Wik · Lund University Publications Student Papers (Lund University) · 2019
Anonymizing time-series data is a demanding task since there might be a lot of private information which can be inferred which might not initially be obvious.Therefore, it can be desired to anonymize with some established privacy guarantee.However, there is always a trade-off to be made between privacy and minimization of information loss.In this thesis, an investigation is made into how to protect univariate time-series.The main focus is on publishing anonymized time-series from individual users, but methods for anonymizing aggregate time-series and the removal of sensitive data is also investigated.This is done in order to find a wider understanding of how a blood glucose related database can be anonymized.The main achievement of this thesis is the implementation of PC-KAPRA, a novel extension to the KAPRA algorithm, which publishes data under (k, P)-anonymity.The results show that PC-KAPRA offers a large improvement in retaining pattern information compared to KAPRA, and publishes data which could be considered qualitative useful information.PC-KAPRA, will be show-cased on other kinds of univariate time-series.Since this is not a thesis in law, there will not be any specific analysis into how the anonymization fulfills the requirements from GDPR.Rather, the focus will be on privatization and specifically anonymization of univariate time-series data from a more algorithmic perspective.