Joe,
I have been lurking on this discussion group because, other than the first
call, I have been unable to attend the calls dues to conflicts.
I understand your confusion around terminology and thank you for your email
trying to clear it up.
I think I follow your logic and, to the extent that I do, I like what
you’re saying.
I do have the following comment.
You identify from ISO 27460-1 “*identity - *set of attributes related to
an entity”
Later in your email you propose the following:
*profile* - set of attributes in a system related to an known entity
*known entity - *an entity with an associated set of attributes in a system
*candidate entity - *an entity who may or may not have a profile in a system
To me the definition of *profile* bears a significant likeness to the ISO
definition of *identity* with the exception that the attributes are related
to a known entity rather than to just an entity (whether known or unknown).
Accepting my premise would change the definitions of the terms you propose
to the following:
*profile* - the identity of an known entity
*known entity - *an entity with an identity in a system
*candidate entity - *an entity who may or may not have a profile in a system
Looking at these new definitions I’d suggest that the term *profile* is, in
essence, redundant to the term *known entity.*
If the terms are the same then I’d suggest that the definitions of these
terms become:
*known entity - *an entity with an identity in a system
*candidate entity - *an entity who may or may not have an *identity* in a
system
Similarly, I’d also suggest that the definitions of the following terms
change as well:
- *identity verification* is the process of relating attributes to a *known
entity* (and recording that correlation in an *identity*)
- *identity proofing* is the process of determining if a *candidate
entity* is the entity represented by an *identity* (in essence, changing
the status of an entity from unknown to known)
- *identity assurance* is the documented process assuring that identity
proofing occurred according to policy
My suggestions are also made with the knowledge that *profile* is an
established term that is already used by Kantara to define the extensions
to a published set of Service Assessment Criteria specified by a federation
to meet its specific requirements. While it is possible to change this term
within the Kantara Identity Assurance Framework, I’d really like to avoid
having to do so.
Thoughts,
Ken
Chair Kantara Initiative Identity Assurance Working Group (IAWG)
On Tue, Jan 29, 2019 at 11:06 PM Joe Andrieu
On the calls I've mentioned a few times my confusion over some of the terminology in the referenced ISO docs. On last week's call, at least some of that was cleared up. This email is an attempt to document my current understanding and any remaining confusion.
At the core of this is the definition of the term identity in ISO 27460-1, as presented in the Notes on ISO Terminology https://kantarainitiative.org/confluence/display/IDPVUseCases/Notes+on+ISO+t... page, and its relationship to the replacement rule.
*identity* set of attributes related to an entity
That same page presents the replacement rule in the following way:
ISO definitions use the 'replacement rule' approach - this means that wherever a defined term appears in text, the reader can directly substitute the definition and the resulting text shall make sense.
Or, in ISO Directives-speak: "The definition shall be written in such a form that it can replace the term in its context."
First, what we cleared up last week was that that when multi-word phrases are defined explicitly, the individual terms are not themselves subject to the replacement rule.
For example, with regards to the term "identity information" on that same set of definitions from ISO 27460-1:
*identity information* set of values of attributes optionally with any associated metadata in an identity
That initial term does not get replaced, i.e., the following would be incorrect application of the replacement rule:
*set of attributes related to an entity information* set of values of attributes optionally with any associated metadata in a set of attributes related to an entity
That part helped a LOT.
However, that also suggest the correct replacement would be:
*identity information* set of values of attributes optionally with any associated metadata in a *set of attributes related to an entity*
Ok. That's awkward, but that's why we have replacement rules, to simplify awkward phrasing. I suppose this is reasonably restated without substantive change as
*identity information *set of values of attributes (and any associated metadata) related to an entity
I can wrap my head around that with the exception that it seems absolutely redundant. Since meta-data *are* attributes related to an entity (albeit with some indirection), both "identity" and "identity information" are the set of attributes related to an entity. As near as I can read, there is no functional distinction between the "value of attributes with any associated meta-data" and the "attributes related to an entity". What are those attributes if they aren't the value? Since meta-data is data, and data that ultimately links back to an entity is related to that entity, is there any distinction here?
Ok. So that's one set of confusion. There seems to be no difference between the meaning of "identity" and "identity information". It's worth noting that the rest of the terms defined (as from ISO 27460-1) use the phrase "identity information" which at least has a certain consistency. And within that set of definitions makes the synonymity a minor issue.
The only exception in that glossary is the term "identity register" used in the definition of "identity registration". That would be replaced as "set-of-attributes-related-to-an-entity registry", which while awkward, at least is clear. I added the hyphens to make it clear that the entire hyphenated phrase is an adjective of registry.
There is also a bit of awkwardness in the definition of authentication, but which I'm ommitting to get to the point that's relevant to our conversation on identity assurance use cases.
Once we get to the definitions for Identity Assurance, I get lost.
To wit (from the wiki in the section started by "SC 27/WG 5 describes Identity Assurance as") :
1.a An identity is established through verification of a set of identity attributes using acceptable evidence or validated systemically against an authoritative data source; then
Is replaced as
A set of attributes related to an entity is established through verification of a set of identity attributes using acceptable evidence or validated systemically against an authoritative data source; then
This makes sense, but the next does not:
This identity is bound to the entity.
Gets replaced as
This set-of-attributes-related-to-an-entity is bound to the entity.
On the surface, this is meaningless to me. I have a conjecture about what it means, but it depends on some adjustments to the terminology and the recognition that these two terms "entity" are not the same in an important way.
My attempt at clarifying this also depends on some subtlety that was suggested on the call last week, but which I did not believe wass consistent with the definition of identity assurance as presented, but that problem might be because the term "assured process" is not defined and is unfamiliar to me.
My question to the group was "what are the differences between identity assurance and identity proofing?" Because the IDPVUseCases wiki page calls out assurance as distinct from identity proofing and identity verification but doesn't explain how it differs. I felt confident in my understanding of verification (as well as how authentication is yet again something altogether different), but I was struggling to understand the difference between assurance and proofing.
The answer that came out of that conversation is that proofing is the process by which you discern whether or not the current user *is* the person you think they are and that assurance is how to prove that the proofing process occurred as defined. This resonated with the distinctions I understand from ISO9001 about quality as not just what you measure and what those measurements should be, but how you document that the measurements were taken and that they met (or didn't meet) the documented requirements. If "assured process" maps to this understanding, than I think I'm on the same page.
Aside from the confusing language, I think there is a conflation here that deserves fixing.
To describe it I'm going to use some alternate language, which after discussion I'm hoping we can map back to standard terms with minimal loss of meaning and minimal desired changes to those standard terms.
*profile* set of attributes in a system related to an known entity *known entity *an entity with an associated set of attributes in a system *candidate entity *an entity who may or may not have a profile in a system
I clarify "in a system" because when you read the non-glossary section in ISO 27460-1, it's clear that the definition of identity *only* applies to the attributes related to an entity *within* a given ICT system. I believe the term *identity* in that specification is better labeled as "profile" with the inclusion of the scoping to a given system.
Using these terms, we can now say, in a much more understandable way that
- *identity verification* is the process of relating attributes to a known entity (and recording that correlation in a profile) - *identity proofing* is the process of determining if a candidate entity is the entity represented by a profile - *identity assurance* is the documented process assuring that identity proofing occurred according to policy
These definitions feel coherent and consistent with what I believe we are talking about. In particular, the term candidate came from Jack Callahan's biometrics work where the Candidate Biometric Vector is matched against an Initial Biometric Vector created at enrollment. His distinction helped me see the conflation of "entity" that was tripping me up in the replaced phrase
This set-of-attributes-related-to-an-entity is bound to the entity.
This would be clearer if the two sentences said
1.a A profile is established through verification of a set of identity attributes using acceptable evidence or validated systemically against an authoritative data source; then
This profile is bound to the candidate entity.
After this step, the candidate entity is now a known entity.
I realize we may not have much influence to adjust the terminology upstream--and that even if this clarification resonates with the people behind those standards it could take a while to incorporate these comments.
HOWEVER, can we clarify for the work we are doing together what is meant by assurance v proofing v verification?
To wit, the use cases I'm sifting through as possible contributions would need slightly different edits for the notion of assurance I gave above so I'd like to make sure we're talking about the same thing.
-j
-- Joe Andrieu, PMP joe@legreq.com LEGENDARY REQUIREMENTS +1(805)705-8651 Do what matters. http://legreq.com http://www.legendaryrequirements.com
_______________________________________________ DG-IDPVUseCases mailing list DG-IDPVUseCases@kantarainitiative.org https://kantarainitiative.org/mailman/listinfo/dg-idpvusecases
-- Kenneth Dagg Independent Consultant Identification and Authentication 613-825-2091 kendaggtbs@gmail.com