Private and Efficient Distributed Learning (PeerL)
The main focus in PeerL is to develop novel private and efficient distributed learning solutions. Distributed learning has several important use cases, e.g.,
within healthcare where it enables hospitals to collaboratively analyze sensitive patient data from individual patients’ wearables or health records without sharing the data, improving diagnostic models and providing personalized treatment recommendations while preserving patient privacy. Another important use case of special interest in PeerL is within insurance where companies can collaborate to train improved crime prevention systems while maintaining the confidentiality of sensitive information.
Of special interest for this use case is a measure of security against leakage of competitive insights, referred to as competitive privacy, a novel concept introduced in this project to quantify the leakage of competitive insights by the sharing of data. The fear of leaking competitive insights is currently considered the main blocker to the wide-spread practical deployment of collaborative learning methods in insurance companies, as well as in the financial sector more broadly, and results from this project will help removing this barrier.
Funding
The project is funded with 12 million NOK by the Research Council of Norway under the Research project for renewal and development of ICT call.