On the initiation of this working group, but prior to the first meeting, Walter sent an email titled "Kicking Things Off" to the group mailing list.
The email asked:
1. From both an IdP and SP perspective, what have been the most common challenges you have encountered when attempting to interoperate with federated partners?
2. To what extent were the challenges from question one a result of operational practices of the site vs. software configuration vs. limitations in the SAML implementation used?
This table (you're welcome, Scott!) is an attempt to capture the issues listed in that thread.
Column 1 captures the identified issues.
Column 2 attempts to recast each issue as a "requirement" (note, the recasting may not work, so this column should be looked at skeptically).
Column 3 categorizes the issue per Walter's note.
Column 4 is for record keeping to identify whether/where each issue is captured and addressed in the work put forward by the working group.
This list will also be used ongoing as a "parking lot" for any issues that are raised in discussion that are not immediately captured in the profile documentation.
| # | Issue | Issue restated as requirement | Limitation | Relevant Profile Sections | Resolved |
|---|---|---|---|---|---|
| 1 | Manual exchange of metadata or (worse) raw config into | Automated, ongoing metadata exchange and validation | Software/Operational | IIP-MD04, IIP-ME04 | Yes |
| 2 | Security risk/change control risk inherent in one-time MD exchange | Automated, ongoing metadata exchange and validation | Operational | IIP-ME03, IIP-ME04 | Yes |
| 3 | Lack of precise documentation and sloppy use of SAML constructs (in custom deployments) | More specificity for use of some specific SAML features | Software | Throughout | Yes |
| 4 | SP-initiated SSO as a "special" case | Support for SP-initiated SSO | Software | IIP-SSO01 | Yes |
| 5 | Lack of deep link support | Support for deep linking | Software/Operational | IIP-SP13 | Yes |
| 6 | Use of frames that break with 3rd party cookies | Keeping authentication screens as top level windows (not iframes) | Operational | Not addressed | |
| 7 | Lack of dynamic provisioning/entitlement-like attribute based authZ | Support for attributes indicating group membership/entitlements (when customers handle authZ) | Software/Operational | Not addressed | |
| 8 | Lack of focus on AuthZ space and support | As above? | Operational | Not addressed | |
| 9 | Lack of clock skew allowance | Support for clock skew | Software | IIP-G01, also recommend adding recommendation for consumption of time server service in a deployment profile | Yes |
| 10 | Lack of encryption support | Support for XML encryption at the SP | Software | IIP-SP13, IIP-SSO04, IIP-MD09, IIP-SP02,IIP-MD10, IIP-MD11, Section 2.5 (IIP-ALG01 - 06), IIP-IDP11, IIP-IDP19 | Yes |
| 11 | Lack of key rollover support | Support for key rollover | Software | Section 2.1.3 (IIP-MD07, IIP-MD08, IIP-SP13, IIP-IDP19) | Yes |
| 12 | Requiring valid (vendor signed and/or expiring) certs | Support for long-lived, self-signed certs, which may or may not be expired | Software/Operational | IIP-MD05, IIP-MD03, IIP-MD11 | Yes |
| 13 | Lack of discovery support/portable links (w/o hard coded IdP refs) | Support for discovery services | Software | IIP-SP09 | Yes |
| 14 | Hard coded 1:1 SP:IdP models | Support for multiple IdPs | Software/Operational | Not addressed | |
| 15 | Require non-opaque, non-transient NameID (rather than attribute) | Support for account identifiers in attributes (rather than NameIDs) | Software/Operational | IIP-SP03, IIP-SP08, IIP-IDP12, IIP-SSO05 | Partial; SP requirements simply state "don't misuse persistent" and "don't require nameid policy in AuthRequests". IdP says "don't require NameID in assertion". Do we need statement about SP accepting assertions not containing NameIDs? |
| 16 | Requiring literal account IDs be asserted by IdP | Support for identifier mapping (i.e., IdP ID is mapped to an internal account ID) | Operational | IIP-SP03 | Best Effort: Whether an SP actually supports this is a configuration issue, agreed that the profile allows for the desired configuration, even if a deployment forgoes leveraging the configuration capability. |
| 17 | AuthnContextClass: not specifying at SP, but failing if PPT not used by IdP | Specify ACC; if unspecified, accept any ACC | Software | IIP-IDP10 | Partial; Addresses the requirement in a roundabout way. Does not state "must not require an ACC if it is not specified in metadata". (Not clear that such a requirement would belong in this document, though). |
| 18 | AuthnContextClass: can't handle locally defined AuthnContextClasses | Allow support of extended ACC's (as part of site-specific configuration) | Software | Possibly; arguably inferable from IIP-IDP10, but it is not clear from IDP10 that IdP must support arbitrary values for ACC. | |
| 19 | AuthnContextClass: no "step-up" support | Support use of "step-up" authentication (re-auth with new ACC and poss ForceAuthn | Software/Operation | Not addressed | |
| 20 | Assuming Logout URL exists | Verify advertised IdP SLO endpoint before directing user there | Software | Section 4.5 (IIP-IDP17-20) | Partial; Says IdP must support SLO, but does not indicate that SPs must honor IdP metadata. Do we need an SP requirement here? |
| 21 | Logoff handling | ??? | SAML | Section 4.5 (IIP-IDP17-20) | Probably |
| 22 | Expectations of SLO | ??? | Operational | Section 4.5 (IIP-IDP17-20) | Partial; (assuming this is largely a duplicate of issue 20) |
| 23 | Browser cookie behavior impacting functionality (sessions not clearing, etc) | ??? | SAML | Not addressed | |
| 24 | Attribute release standards for IdPs | ??? | Operational | Not addressed | |
| 25 | Attribute release: suppressing grad students (FERPA concerns) | ??? | Operational | Not addressed | Is this and 24 about configuring conditional release of data from specfiic users? |
| 26 | Privacy practices: what is actually being kept private? | ??? | Tangential | Not addressed | |
| 27 | Standardized and effective workflow for dealing with attribute release | ??? | Operational | IIP-IDP05, IIP-IDP06, arguably IIP-MD04 | Partial; IIP-IDP05 is useful for support of entity categories, and IIP-IDP06 is useful to the extent that including md:RequestedAttributes is part of the operational solution. IIP-MD04 is useful to the extent that consuming or excluding metadata simplifies the process |
| 28 | Vendors charging fees for setup and support of SAML | SAML support should be part of base service | Operational | Out of profile scope | |
| 29 | Lack of framework/contract terms; change controls, support escalation | ??? | Operational | Out of profile scope | |
| 30 | Lack of testing SP/IdP facilities (test SP, test IdP) | Run a testing SP/IdP for validation purposes during initial integration testing? | Operational | Not addressed | |
| 31 | Knowledge gaps with some vendors on how SAML works. | ??? | Operational | Out of profile scope or The entire document | |
| 32 | Advertised but unsupported functionality in metadata (artifact endpoints, etc.) | Advertise only supported endpoints | Operational | IIP-MD09; IIP-SP02; IIP-IDP02 | Partial; MA01-02 address listed encryption profiles. Arguably the metadata exchange requirements imply some support of this, but no specific requirements are listed. |
| 33 | Availability of POP/mechanism for assessing risk | InCommon: stronger focus on POP? [May be addressed in different workgroups] | Operational | Out of profile scope | |
| 34 | Publishing metadata contact info for security incident response | Include security incident response (usually security or help desk) in metadata | Operational | Out of profile scope | |
| 35 | ForceAuthn: IdPs not ensuring user is reauthenticated | Verify function of reauth before resetting authninstant | Operational | IIP-IDP08 | Yes; at least to the extent we can define it across authN methods. |
| 36 | ForceAuthn: SPs not checking authninstant | Verify (or allow verification) of authninstant currency | Software/Operational | Not addressed | |
| 37 | OASIS Standards have not been updated with Errata, current Errata out-of-date | Recommend in report-out of WG that someone be resourced to update the Errata and a modify the standard to include the changes from Errata (working with OASIS) (Scott C says someone has informally volunteered to do this. Who?) | Standards | Out of profile scope | Partial; Addressed separately (Scott C, Eric), but not included in the OASIS repository. |
| 38 | Review with REFEDS once a solid draft is done | Nick to check in with Nicole on this | Standards | Out of profile scope | Nick |
| 39 | Research collaboration requirements for adoption of a persistent nameID | Use of persistent nameID or other mechanism to enable seamless collaboration across multiple SPs in a research organization. | Operational | Out of profile scope? | Scott K |
| 40 | "Ready For Collaboration" entity category for IdPs | Description of an entity category that would signal that an IdP is configured for ease of collaboration with no manual intervention by operators, does not re-assign ePPN, and/or uses persistent nameID... etc. TBD | Operational | Out of profile scope? | David W |
| 41 | "Red IdPs" | eduGAIN has the "ECCS" service (https://technical.edugain.org/eccs/index.html) for highlighting various levels of IdP operability. Tom Scavo has a script that looks for "dead" IdPs. Is there some useful baseline for IdP operability or interoperability that this group would recommend and could it be tested for? | Operational | Out of profile scope, in scope for later work of a successor to this group | Nick / Scott Koranda |
| 42 | Don't respond to Unsolicited assertions. | (Still working to clarify specific requirement) | Software | Not addressed | |
| 43 | Include language in SAML2int regarding support for multiple IdPs asserting against access to the same resource URL/entityID. (I.e., clarify that federation presumes cloud vendors can support multiple IdPs and discovery, not just externalized authentication) | Software/Operational | Followup to item 14 to be addressed in SAML2INT work | ||
| 44 | Attribute or NameID values too short or disallow legal XML characters | Minimum implementation requirements for attribute/nameid values (in particular xs:string) length and legal characters | Software | ||
| 45 | Lack of scope validation | Attribute scopes can be validated against allowed scopes defined in metadata (or elsewhere?). | Software. | ||
| 46 | Lack of time synchronization (separate from, but as important as clockskew) | Require that SP and IdP deployments use time synchronization against time servers | Operational | Not addressed | |
| 47 | Java and md5/sha1 certificate support | Deployment profile should call out that all certs should be signed with modern signing algorithms to avoid being rejected by cryptographic code that is increasingly aggressive about rejecting older signature types, even in cases where signature verification is not required. | Operational | Not addressed | |
| 48 | Binding of an identifier to its issuer or more broadly checking scope | See: http://www.economyofmechanism.com/office365-authbypass.html Issues that need to be remedied in a rev of saml2int: require binding of an asserted identifier to the public key in metadata of its original issuer. | Software/Operational | Not addressed |
Note: not included here are some recommended reference links, as those have been captured in the working group's list of references already