The wiki is deprecated and due to be decommissioned by the end of September 2022.
The content is being migrated to other supports, new updates will be ignored and lost.
If needed you can get in touch with EGI SDIS team using operations @ egi.eu.
GOCDB/notifications
| Main | EGI.eu operations services | Support | Documentation | Tools | Activities | Performance | Technology | Catch-all Services | Resource Allocation | Security |
| Tools menu: | • Main page | • Instructions for developers | • AAI Proxy | • Accounting Portal | • Accounting Repository | • AppDB | • ARGO | • GGUS | • GOCDB |
| • Message brokers | • Licenses | • OTAGs | • Operations Portal | • Perun | • EGI Collaboration tools | • LToS | • EGI Workload Manager |
Use of eppn in EGI
Data Privacy and Code of Conduct Policy Document Under Construction: GOCDB/data_privacy
Hi Ops/all,
Apologies for the lengthy text:
With the plans to include federated identity, services like GocDB (which falls under the R&S REFEDS SP Category) will need to use an attribute supplied by an IdP to create/identify a user account, usually ‘eduPersonPrincipleName’ (eppn) (or less likely ‘eduPersonTargetedID’, eptid, see discussion below). These attributes will be used instead of an x509 DN and will require changes to the output of the PI, changing CERTDN element to PRINCIPAL element in queries like get_user and get_site_contacts - GocDB needs to republish these attributes in its PI for use by the rest of the infrastructure (e.g. queried by Ops portal, used in accounting and so on).
Now, according to the rules of the UK Access Fed (similar rules apply in other feds), “The Service Provider must not disclose to third parties any Attributes other than to any data processor of the service provider or where the relevant end user has give its prior informed consent to such disclosure. see [note2]” . (Where [note2] states: “The basic Rule is that attributes may only be used by the service requested by the user and only for the specified purposes. Service Providers that wish to use attributes in other ways (for example to provide direct user support) can arrange this either by obtaining positive informed consent from each individual End User, or by contract with Identity Providers who are then responsible for informing their End User.”)
Therefore, to use Attributes like the eppn in EGI we need to obtain ‘positive informed consent’ from the user that we can publish and make visible their data/eppn to authenticated users and client-services of GOCDB, including those trusted by UK Access Management Federation, EGI-SSO or IGTF (I think we must explicitly state the different authentication/trust realms, e.g. any person/host with a valid IGTF cert can query the GocDB API for this info).
- Q. Any experts out there know if the sample notification/approval-dialog shown in the screen grab below provides adequate positive consent?
- Update, looking at: [on the definition of consent]
- Q. Assuming user clicks OK, do you know if there are any issues for EGI and for the hosting institution (STFC) in re-publishing the eppn in the PI? (it then becomes readable to any other service/user who has a valid IGTF cert)
- Update 20/08/2015: Re-publishing the eppn looks like a no-go: The SP needs to first take account of each/every IdP's policy on whether the eppn can be reused. This is does not scale. However, hashing the ePPN or using the ePUID may be ok (see below).
Using ePTID without ePPN (not suitable)
GocDB could use the ‘eduPersonTargetedID’ (eptid) and re-publish that in the PI – the eptid is less of a security concern because its anonymous/opaque. However, since the eptid will be different for the same user across different SAML Service Providers, this provides little use when there is requirement to identify a particular accountID across different SPs and organisations (this is the EGI requirement). Note, if EGI had a single centralised/proxy IdP, it could be used to re-issue the same eptid consistently to all the underlying EGI services/systems. I think this would be far preferable compared to services individually registering with federations and receiving different eptids per service.
- Update 20/08/2015, using the eptid is not suitable as it does not match the EGI requirement
Proposed Solution
- The proposed solution makes use of a a) secret salt value shared amongst project SPs to hash the ePPN, and b) uses the ePTID to solely determine if an ePPN has been re-allocated:
a) Hashing the ePPN using a project-owned secret salt; a way to securely re-publish and correlate attributes/IDs within a project
A group of related Service Providers, e.g. those belonging to a particular project, may concatenate a secret ‘salt’ value to the value of an attribute before hashing the result, for example; hash(‘secretSaltValue + eduPersonPrincipleName’). In doing this, the same hash value can only be reproduced if: a) an SP has access to the same secret salt value and, b) the SPs agree to use the same hashing algorithm. If the salted and hashed attribute is the eduPersonPrincipalName, this produces an opaque and persistent identifier that can only be reproduced and correlated across the SPs within the project (i.e. those that have access to the salt). Provided the salt value is not disclosed to any third parties, the resulting hash string can be safely republished, for example, for use in correlating account IDs across project-specific SPs in the same project.
b) Requesting both ePPN and ePTID to determine ePPN re-allocation
The ePPN is re-assignable, and re-assignment practices depend on the federation/country; you cannot always rely on a federation not re-assigning an ePPN after a period of inactivity (e.g. 2yrs with the UKAMF). Therefore, as suggested by the R&S REFEDS SP Category, its good practice to also request/release both ePPN and ePTID together. This is useful because the SP can also record an ePTID change (by definition the ePTID can never be re-assigned). By doing this, the SP knows that the holder of the ePPN has changed. Note, the ePTID would only be used internally by a particular SP to check that the user has not changed; it wouldn't be used when hashing the ePPN with a secret salt for use in republishing as described above.
- Q. I think this could be used by SPs within EGI allowing accounts to be safely correlated across SPs.
I’d be grateful for any thoughts/guidance on these matters (many thanks to Mikael Linden and Andrew Cormack for their help).
Cheers,
David
