Solution architecture

Target architecture for mediam

Instead we propose to implement a zero trust architecture (which we define as the ability to manage identity and authorization at every step), relatively similar to the one proposed by ockam, but with the following changes :

• our identity and access management layer is centered around IETF GNAP and DIF KERI: to our knowledge this is the first integrated that bridges advanced standards on authorization and delegation, together with decentralized identities because we need to be able to work seamlessly between different organizations and with external patients, so we can't only rely on a silo-ed identity registry (while still being compatible with existing OAuth2/OpenID connect to ease adoption)

• vaults are extended to support confidential computing techniques (we made several prototypes during medIAM, for instance the capability to make private artificial intelligence between 2 healthcare organizations, without releaving the raw data itself)

• an opiniated difference is that HTTP is our main transport layer, albeit with a focus on HTTP/3 to enable IoT scenarios. Interoperability with non-IP layers is made thanks to SCHC compression (and has been demonstrated with LoRa), allowing convergence between IoT and cloud computing. We may later provide bridges to ockam as well, as it provides more IoT specific features such as enrollment (but within medIAM we only tested it separately). The integration is made easy since both projects are implemented in rust and opensource.

Proposed medIAM architecture

Those choices take into account that most healthcare organisations can't deal with IoT security separately. As discussed in the introduction, a solution must deal with IT, biomedical devices and building management. The only way to implement zero trust is to protect all resources through a common protocol (in our case IETF GNAP).