For many application domains benchmarking is relevant, specifically to compare the position of an organisation (e.g. a health institute, bank, ZZP’er) with that of competing organisations. Sometimes trusted third parties may be used for specific benchmarks (e.g. ECB), but these types of benchmarks are often very generic, and do not take into account specific relevant attributes, and such a TTP can be very expensive. Also, specific kinds of data may be very sensitive to share (commercially confidential as well as privacy-sensitive). It is therefore interesting to investigate to what extent MPC and secure learning may be used for confidential benchmarking without the need of a trusted third party. In addition, it would be interesting to make the benchmarking “actionable”: can we learn (from sensitive data) which attributes contribute most to the benchmark position of an organisation, such that an organisation may take this into account to improve their position in the future?
We aim to develop generic confidential benchmarking solutions using privacy preserving analytics technology, in particular multi party computation. Initially this will be domain-independent. Currently the idea is to work towards a “confidential benchmarking” demo for one of more application domains, including benchmarking in the health domain.
Click here for more information