Workshop on privacy-preserving statistical computation with Statistics Norway


Report Abstract

Based on the presentations by the speakers from the Cancer Registry of Norway, Statistics of Norway and Norwegian Center for E-health Research, as well as the number of questions raised up from the audience, we can conclude the problem of privacy preserving to be relevant both in healthcare and other domains where statistical analysis of data combined from several sources is performed.

Different solutions can be applied for this issue. Pseudonymization and de-identification techniques still leave the space for sneaking into the individuals’ data due to more information available in the joined databases; linking additional publicly available information about the individuals (for example, from social media) enhances the privacy risk.

Secure multi-party computation techniques based on blind data miners can become a solution for preserving privacy in statistical studies with data from several sources. With SMC, all the required computations are performed without revealing any microdata to the computing entities: each party learns only the corresponding function output value and no inputs of other parties. The blind data miners run secure protocols that compute statistical functions on the data producing the aggregated results. This is especially useful when the source databases are confidential and should not be openly linked. Herewith, both individuals’ and health institutions’ privacy can be protected saving the quality of research results. Additionally, costs of conducting statistical studies on distributed data can be reduced.