D7.2 Client connectors' architecture design – Final version

The client connector is the component required for a participant to join the MUSKETEER Platform. It is the software application supporting MUSKETEER platform participant in his data exchange, share and process, so to guarantee that he is sovereign of his data. The client-side connectors have to support the set of privacy operation modes (POMs) made available throughout the project according to the architecture defined in T3.1 and meet the requirements of the federated and privacy-preserving machine learning services designed in WP4. Moreover, the client component provides services for locally combining model updates into one consistent, up-to-date model instance. The client component serves as adaptor for the integration and industrial validation of the MUSKETEER platform in WP7. The client connector, running on a secure and private space, provides the interface (Client Connector External APIs) for receiving a set of instructions from a master controller of the MUSKETEER server, related to the transferring of the required datasets and/or models (according to the POM chosen) from/to the MUSKETEER core to the secure and private space for the training of an MUSKETEER ML model