| Popis: |
In February 2024, Israel's Ministry of Health released microdata of live births in Israel in 2014. The dataset is based on Israel's National Registry of Live Births and offers substantial value in multiple areas, such as scientific research and policy-making. At the same time, the data was processed so as to protect the privacy of 2014's mothers and newborns. The release was co-designed by the authors together with stakeholders from both inside and outside the Ministry of Health. This paper presents the methodology used to obtain that release. It also describes the considerations involved in choosing the methodology and the process followed. We used differential privacy as our formal measure of the privacy loss incurred by the released dataset. More concretely, we prove that the released dataset is differentially private with privacy loss budget \varepsilon = 9.98. We extensively used the private selection algorithm of Liu and Talwar (STOC 2019) to bundle together multiple steps such as data transformation, model generation algorithm, hyperparameter selection, and evaluation. The model generation algorithm selected was PrivBayes (Zhang et al., SIGMOD 2014). The evaluation was based on a list of acceptance criteria, which were also disclosed only approximately so as to provide an overall differential privacy guarantee. We also discuss concrete challenges and barriers that appear relevant to the next steps of this pilot project, as well as to future differentially private releases. |
