This is a call to Join the running MPC use cases on Fraud Detection and Combatting Poverty, or join the ideation on Secure, distributed learning.
Within Techruption, we investigate applications of technologies for privacy-proof data analytics: Multi-Party Computation (MPC), and related techniques like federated learning, enable organizations to jointly analyze (and learn from) their sensitive data without having to share or reveal this data. See www.tno.nl/mpc for more information. Below more information on of the two running use cases (collaborative fraud detection and combatting poverty) and an upcoming use case on secure, distributed machine learning. For all three, also new partners are welcome!
Why and What?
Poverty is a challenging problem in many municipalities. There is a strong motivation to detect poverty early, or even prevent it. This is relevant for the municipality of Heerlen, which often ranks high on poverty lists, but also a broader societal challenge. Generally, the approach to tackle poverty is after-the-fact: it starts when someone is already in poverty. The municipality would like to turn this into a pro-active approach; however, there is too little insight into poverty and potentially related (and predictive) other factors such as payment problems, the use of utilities, health costs, income and demographic factors. One of the key obstacles to analyzing these relations is that the underlying data is privacy-sensitive and cannot be shared. This use-case, therefore, explores the extent to which the municipal government could use Multi-Party Computation to gain relevant insights into the different factors influencing poverty without revealing or sharing data at the individual level, preserving the privacy of individuals.
In 2020, TNO, CZ, CBS, municipality of Heerlen and University of Maastricht will develop a Proof-of-Concept, including a dashboard, to enable performing relevant queries on combined poverty-related data, investigate functional and non-functional requirements, and investigate legal constraints.
This use case is interesting for organizations that wish to contribute to combatting poverty and have (potentially) relevant data that may be related to poverty issues. Particularly: municipalities, data providers.