Consortium between competitors
Machine learning algorithms require a large amount of data to increase their performance. But, within an industry, this data is often scattered among different actors who cannot afford to share what is the core of their business.
HOW TO INCREASE THE PERFORMANCE OF AN ENTIRE INDUSTRY THROUGH MACHINE LEARNING, WITHOUT COMPROMISING THE SECURITY OF EACH COMPETITOR'S DATA?
A new model of collaboration
Develop efficient machine learning models based on data from several companies and partner organizations, without exposing the relevant datasets, and thus allowing participants to preserve their core business values.
PROTECT THE DATA
Sensitive data remains on the infrastructure specific to each consortium member. Only metadata and models are shared between the different partners.
Each partner preserves the value of its data.
ENSURE TRACEABILITY
Based on Hyperledger Fabric, the distributed ledger within Substra allows to manage the rights and authorizations of each partner and to track all actions performed within the consortium.
No action can be carried out without the agreement of the consortium participants.
CO-BUILD MODELS
Thanks to transfer learning, partners can decide to share only part of the model, the lower layers constituting a common core, the upper layers remaining private.
Only a common core can be shared to strengthen data confidentiality.
Reinventing your ecosystem
The Substra framework allows to create new partnerships around machine learning while guaranteeing data security, benefiting all participants.
You want to know more about it? Discover how the MELLODDY project brings together a consortium of 17 partners in the European healthcare industry to accelerate the discovery of new molecules.