...
- Numerous references to "digital service" without being clear what they mean.
- Refer extensively to the "subscriber" whereas other schemes include "subscriber" and a "subject". The "subscriber" may be the organization who wants credentials issued to a number of subjects. Would make easier alignment with other sources.
- Model and state diagrams are inconsistently applied. When you become a subscriber because you have enrolled, the definition of enrollment doesn't include the definition of a service account. Since they are unclear on the enrollment process its not clear what the subscriber is.
- RGW suggests maybe this means that 800-63A should include the idea of enrollment and becoming a subscriber/subject. Overall, changing the term from subscriber to subject.
- Andrew is wrestling with the question of "are you still a subscriber when you are federating authentication?"
- Part of the model inconsistency is differences in how the verb "authenticate" is applied, does the subscriber authenticate or does the CSP do the authenticating.
- Andrew is reviewing and finding internal inconsistencies in the way terms are used, it's not clear what the state model is to get from non-authenticated to authenticate state, versus the authenticators and secrets and other things needed to assert the identifier. They do say that the purpose of authentication is to produce an identifier, versus the purpose is to get access to a service.
- Scott suggests whether it's possible to use the term identified access to a service.
- Andrew notes that they reference "classic kerberos" versus "modern kerberos"
- Denny notes that there's a section around usability, but how it looks on the screen and the user interface, is that something that is in the scope now? Andrew Hughes responds that they obtain usability from following the NSTIC guiding principles, which include usability. The idea being that authenticators that are difficult to use are not trusted.
- Ken notes that there are currently no normative usability requirements.
- RGW mentions the issue of uniqueness of credentials, and inconsistency in the use of the term "digital service" in the introduction to SP-800-63-B. Section 4 uses the term as if it refers to the Credential Service Provider, whereas the introductory text uses the term as if means the Relying Party.
- Section 4 is marked normative, so it should be clear about requirements, use shall statements for that.
- On guidance versus technical requirements. Richard observes that on the first page they should be referring to requirements rather than guidance. Calling it guidelines dilutes the force of the requirement. Andrew can observe that they are putting normative statements in the document but not using normative language. The documents are called "guidance" but it contains requirements. Are the contents mandatory or not?
- Andrew observes that a state model would be helpful to show how entities go from non-authenticated to authenticated state, verifiers in pre-authenticated to post-authenticated state. Scott concurs on the state model.
- Denny asks about if there's a model for them to reference. Andrew observes that 800-63-3 describes the model of the architecture for the discussion in the documents. There's a role diagram that infers some changes of state, it's indistinct but it illustrates that a claimant becomes a subscriber. Probably not complex, but not documented at this point. Denny wonders about disability acts in various parts of the world.
- Additional comments from RGW - section 4.1.2 - cryptographic authenticators at AAL1 shall use approved cryptography. Scott suggests that this is in line with NIST's mission to push approved cryptography for all uses of government cryptography.
- Additional comment RGW - 4.1.4, 4.2.4, 4.3.4 there are references to 800-53 "or equivalent standard" but what is the method of judging what an equivalent standard is.
- Andrew has a comment on section 4.5 - summary of requirements, they don't have rows for records retention or privacy requirements.
Reference: IAWG Meeting Minutes 2017-03-09
SIXTH SESSION
Key discussion items March 16th:
REVIEW OF 800-63C
- Credential generation and other lifecycle issues are missing from the discussion. Andrew points out 800-63B has a section called lifecycle management.
- Ken asks if anything changes if it happens in a federated context as opposed to the context B was written in.
- RGW suggests that it depends whether the federation includes requirements to be a member of the club. Only becoming more of a concern as reading 63B and 63C. Many SHOULD statements - as we know, if is says SHOULD then they probably won't.
- Globally we have a comment that SHALL and SHOULD need to be clear. Each distinct SHALL or SHOULD ought to be in a single paragraph.
- Andrew observes it's a similar comment to last week - the document is a mixture of explanatory material, guidance material and requirements material.
- Ken suggest we could comment them for adopting a normative style.
- General agreement that the document is not ready for prime time.
- Andrew notes that we appreciate the shift towards normative language in the requirements, but the phrasing of some requirements makes it difficult to have certainty that the implementation meets those requirements. As assessors there is also uncertainty about how to evaluate the conformity. Uncertainty then leads to inconsistency.
- RGW has one other broad topic - 4.2 of 63C - requirements on federal agencies slapped on the end of the section. Perhaps including it in an annex instead of including in the rest of the flow of the document. The agency guidance at the end of the privacy section is a non-sequitur with respect to the rest of the document
- Andrew notes that the audience section of 63-3 is blank.
- We could use clarity from the authors on when the agency specific text applies.
...
- Next
...
- steps: Thursday 243rd we will take the first cut at looking at the comments. We can package and submit them early if we're happy with them next week.
Reference: DRAFT IAWG Meeting Minutes 2017-03-16