How do the different topologies compare with regard to data protection and privacy?

Question

Accepted Answer

Isolated identities spread the user's data over many providers of uneven quality; central and IdP-centred models concentrate it at one IdP that can profile every login; full mesh keeps identities at the user's home organisation with no global observer; hub-and-spoke puts a broker in the position to see every authentication in the federation; user-centric models take the issuer out of the usage path entirely.

Topology
Who sees what

Isolated
Each RP holds its own copy of identity data and credentials; many small honeypots, no cross-site profile

Central / IdP-centred
The IdP is in every login and can build a cross-RP profile; one large honeypot

RP-centred
The RP sees which IdP each user chose; each IdP sees the logins to that RP

Full mesh
The home IdP sees its own users' logins to partner RPs; no single party sees everything

Hub-and-spoke
The broker sees every authentication and, through its proxy, the content of the assertions

User-centric / SSI
The issuer learns nothing at usage time; only the RP sees what the user chooses to disclose

The rule of thumb: the more a topology centralises convenience, the more it centralises observation. Privacy-preserving designs either avoid the central party or blind it, which is the motivation for the self-sovereign approach.

Topology	Who sees what
Isolated	Each RP holds its own copy of identity data and credentials; many small honeypots, no cross-site profile
Central / IdP-centred	The IdP is in every login and can build a cross-RP profile; one large honeypot
RP-centred	The RP sees which IdP each user chose; each IdP sees the logins to that RP
Full mesh	The home IdP sees its own users' logins to partner RPs; no single party sees everything
Hub-and-spoke	The broker sees every authentication and, through its proxy, the content of the assertions
User-centric / SSI	The issuer learns nothing at usage time; only the RP sees what the user chooses to disclose