[MINUTES] VCWG Confidence Method 2026-08-13

This meeting focused on refining the VCWG Confidence Method specification
through the review and discussion of several pull requests. Key discussions
revolved around clarifying the language of the confidence method,
particularly regarding its application to various subjects within a
Verifiable Credential (VC), and how to incorporate concepts like
cryptographic keys and biometric images. The group also debated the
specification's title and the placement of assurance levels within the
document. The aim is to prepare the specification for horizontal review by
the Technical Architecture Group (TAC).

*Topics Covered:*

   - *Clarifying Confidence Method Language:* The attendees discussed the
   generality of the confidence method's language and how to make it clearer
   for newcomers, with a focus on its application to various subjects within a
   VC, not just the holder. This involved exploring concrete examples like
   marriage licenses to better illustrate the concept.
   - *Pull Request 43 (Cryptographic Keys and Controlled Identifiers):*
   This pull request aimed to add sections on cryptographic keys and
   controlled identifiers, sparking a detailed discussion on the scope of the
   confidence method and its association with different types of subjects
   within a VC, including those that are not direct credential subjects.
   - *Pull Request 45 (Align with Original VCDM 2.0 Definition):* This pull
   request was approved and aims to align the specification with the original
   VCDM 2.0 definition, with the outcome being the alignment of the spec with
   previous definitions.
   - *Biometric Image Section (Pull Request 44):* The discussion centered
   on the scope of biometric images, specifically whether to limit it to
   facial images or to generalize it for other biometrics like irises and
   fingerprints, and ensuring privacy considerations are adequately addressed.
   The group decided to modify the language to allow for extensibility in
   future versions, focusing on facial images for the current specification
   while acknowledging the need for privacy safeguards.
   - *PR 47: Title and Assurance Level:* The attendees discussed proposed
   changes to the specification's title and the placement of assurance levels,
   debating whether assurance levels should be a separate top-level concept or
   a child of confidence methods. The outcome was an agreement to keep
   assurance levels tied to the VC issuance and to continue discussing the
   most appropriate title for the specification.

*Action Items:*

   - Manu Sporny will update Pull Request 43 to incorporate the discussed
   concepts regarding confidence methods, subjects, and concrete examples like
   marriage licenses, and to iterate on the text asynchronously with Ted
   Thibodeau Jr.
   - Denken Chen will review how the preview and diffs work for pull
   requests to ensure they are functioning correctly.
   - Denken Chen will adjust the language in the biometric modality section
   of Pull Request 44 to allow for extensibility beyond facial images and will
   reach out for further input.
   - The editors (Denken Chen and Joe Andrieu) will create a punch list of
   tasks needed to prepare the specification for horizontal review by TAC.

HTML:
https://meet.w3c-ccg.org/archives/w3c-vcwg-confidence-method-2026-08-13.html

Video:
https://meet.w3c-ccg.org/archives/w3c-vcwg-confidence-method-2026-08-13.mp4

[image: W3C] <https://www.w3.org/>
VCWG Confidence Method 13 August 2026 Attendees

Present

Dave Longley, Denken Chen, Elaine Wooton, Joe Andrieu, Manu Sporny, Scott
Jones, Ted Thibodeau Jr

Regrets

-

Chair

-

Scribe

transcriber
Contents

   1. Reviewing Pull Requests <#f748>
   2. Clarifying Confidence Method Language <#eb75>
   3. Pull Request 43 Discussion <#f910>
   4. Add sections on cryptographic keys and controlled identifiers by
   msporny · Pull Request #43 · w3c/vc-confidence-method · GitHub <#256e>
   5. Pull Request 45 Discussion <#fcde>
   6. Align with the original VCDM 2.0 definition by denkeni · Pull Request
   #45 · w3c/vc-confidence-method · GitHub <#0b3b>
   7. Biometric Image Section <#9df7>
   8. Add biometric image section by denkeni · Pull Request #44 ·
   w3c/vc-confidence-method · GitHub <#44b8>
   9. PR 47: Title and Assurance Level <#7908>
   10. Editorial cleanup: sections, title, tooling by denkeni · Pull
   Request #47 · w3c/vc-confidence-method · GitHub <#8ca0>

Meeting minutes Reviewing Pull Requests

Joe Andrieu: Howdy folks. How's it been going without me?

Scott Jones: Hi. …

Scott Jones: the last one was cancelled, I believe.

Joe Andrieu: So it's been going horribly.

Denken Chen: Yeah, we have the conflict and…

Denken Chen: also I haven't got assigned as the host for this online
meeting yet. the test has manual data. Yes.

Joe Andrieu: I'm not sure what you mean about the host tanking.

Joe Andrieu: I think Dankin, you had mentioned you had a couple of PRs
queued up for us.

Denken Chen: Today I think we can go through the four PRs and hopefully you
can get consensus and merge them as early as possible and…

Scott Jones: It's crazy.

Denken Chen: so we can look at the juice and…

Denken Chen: show screen or I can do that as One minute.

Joe Andrieu: If you could that would be appreciated.

Denken Chen: So we have this 43 inserts from manual helpers and…

Scott Jones: speaks about certain things.

Denken Chen: also we review it a little bit and then the last one is raised
by pet and that is available to speak about this a little it. Sure.

Ted Thibodeau Jr: I am not prepared to speak or anything but I can try to
say what I was talking about. as written the sort of genericy text blurs
things even to my mind and…

Denken Chen: Thanks.
Clarifying Confidence Method Language

Ted Thibodeau Jr: I think I'm about as informed as anybody else in the
group. and I think that blurring is going to be problematic for people
coming to it.

Scott Jones> Wow, cryptographic dog collars. Fascinating!

Ted Thibodeau Jr: It is not clear what role is actually being talked about
and part of that is because we have the triple role right sub holder issuer
and verifier and then we've got these other things that happen the subject
of which there may be many the presenter the issue and none those three are
part of the role goal set, but they sort of need to be talked about. And
what we're doing with competence method still doesn't flow to me. So,
that's the part I have the biggest problem. what it is that is where the
confidence is being increased and why is not clear. It does not stay clear.

Ted Thibodeau Jr: We have a conversation, it becomes clearer for a bit and
then we go away and it becomes blurry again and it gets worse with these
generic term segments like the one that I was commenting on there
specifically. And that's what I said in the end is fully generic text and
something more like that dog hypothetical that Manu put in.

Denken Chen: Sure.

Ted Thibodeau Jr: I'll leave it at that for now because I really didn't
plan to speak.

Manu Sporny: Plus one to what Ted's saying. I was struggling with the
language here because confidence method is so kind of generic and
generalized, you can use it on a human being, you can use it on a dog
collar, you can use it on an AI, you can use it on a car. you can, put a
confidence method on just about anything to raise your confidence that it
is the thing that you think it is. And there's so many different types of
confidence methods like a picture of the thing or a cryptographic key tied
to the thing or a facial vector of the thing just a shape vector of the
thing. it's very generalized and it's not necessarily the holder. which is
why

Manu Sporny: the issue which is why I didn't want to kind of put that in
there It's just much more broad. So maybe what we do here is provide a
concrete example that is more believable than a cryptographic dog collar.
and maybe we go from there we use a gener generalized language and then
we're like for example I don't know I can think of something but if that's
the direction the group wants to go to. I do think Ted does have a point.
It's so general that people reading it for the first time are just going to
have a very hard time understanding what's actually being said there.

Manu Sporny: And maybe the concrete example is we do talk about the holder
being the subject and that is the most we expect use case today where
holder shows up they prove a cryptographic binding to a private key and the
issuer puts that public key binding in the credential that they issue. to
the subject which is also the holder and then when the holder shows up at a
verifier they do the same cryptographic binding proof ceremony there. So
maybe that's our concrete example that we use here.

Ted Thibodeau Jr> "cryptographic dog collars" not far removed from
"cryptographic pet implants"

Dave Longley: I was hearing is it all right for me to jump in?

Scott Jones> like a microchip?

Denken Chen: Please. Yeah,…

Dave Longley> confidence methods can be used during presentation to
increase confidence that the presenter can authenticate as or on behalf of
the subjects the confidence methods are for

Denken Chen: no problem. Please.

Ted Thibodeau Jr> yes, like a microchip. My cats are "chipped". Dogs,
parrots, various other "prone to wandering" pets get them.

Dave Longley: I was hearing from Ted that it gets confusing about what
we're increasing the confidence of and I tried to type something into chat
here that might be helpful language. my understanding is that confidence
methods can be used during presentation to increase confidence that the
presenter can authenticate on behalf of the subjects the confidence methods
are for. So I do think we need to focus on when you're using a confidence
method, it's during presentation and you're doing it to increase confidence
that the presenter can do this authentication and it could be as a subject
or on behalf of a subject or multiple subjects in the credential.

Dave Longley: And that really is separate from holders and…

Dave Longley: issuees and so on.

Denken Chen: Yeah, thanks.

Denken Chen: Money, please.

Manu Sporny: I'm wondering if that's correct. because what about theam the
case I think this still happens at presentation but let's say you have a
business registry right and you've got a business listing and it's just
available to the public and there's a confidence method listed against one
of the businesses bound to a cryptographic key and then somebody else wants
to figure out if you are associated you are operating on behalf of that
business. And in that case, it does happen during presentation, but it's
very much out of band. It's not like the business isn't necessarily the
verifier already has the credential that's associated with the business. so
there that's not presented, right?

Manu Sporny: and there's some kind of band ceremony where the verifier is
like, "Hey, prove that you have possession of this key where you don't even
send the credential over during presentation." So, it's like corner cases
like that that I'm concerned where the language kind of breaks Down.

Dave Longley: That did not seem to break anything in my mental model.
there's still a presentation involved and the for that presentation. It
doesn't matter. what the authentication process is that is you are
authenticating and the authentication is related to that presentation not
some other presentation.

Dave Longley: So if there's some phone call about that presentation that
happens later and that's your confidence method, that totally seems like it
fits the

Denken Chen: Yeah, I would like to see from another angle.

Denken Chen: I'm plus one to minus first paragraph mentioned that we don't
use holder in this one. it seems weird at the first glance because in the
VC ecosystem we mentioned it's a third party model is between the issue
holder and verifier.

Denken Chen: However, since the beginning of this confidence back, Jordan
and I decided to focus away from the holder but into the subject
particularly when the subject is not an human entity could be a pet or
multiple subjects. So the whole VC confidence method is really about the
verifier and the subjects on the VCs and so for example when the holder is
one of the subjects that's one use case but that's just one of them.

Denken Chen: So probably it would help that we add some wordings at the
beginning of this spec mentioning that this spec will deal mostly about the
subject and the verifier.

Denken Chen: So we can move away from the thinking of that we are not just
dealing with the holder itself. Yeah, please.
Pull Request 43 Discussion Add sections on cryptographic keys and
controlled identifiers by msporny · Pull Request #43 ·
w3c/vc-confidence-method · GitHub
<https://github.com/w3c/vc-confidence-method/pull/43>

Ted Thibodeau Jr: The trouble with that is that the issuer is intimately
involved…

Ted Thibodeau Jr: because they have to put this verification method into
the VC to start with. That's it.

Denken Chen: I don't see the problem there. But Joe, please go ahead.

Joe Andrieu: Yeah, I didn't see the problem either.

Joe Andrieu: Ted, the issuer definitely sets everything in the VC and they
are the one who is providing this confidence method which is a hint for
something the verifier can do to verify that the subject of this VC is the
same as some other entity in their system. It may be someone who's
authenticating live. It may be someone who invoked a capability
historically and they're looking to say, "Hey, that someone signed this
Zcap over here. Is that the entity that we had a confidence method for in
this VC?" those are all valid use cases and it applies to any RDF subject

Joe Andrieu: in the subject claims. So it's not the notion that there's one
subject of a VC, which was sort of a simplification that let us talk about
this for driver's license. It's easy to think of one subject. if you were
to think of a marriage license, there are probably four or five people who
need to be identified as individual subjects. If you're recreating the ones
I'm familiar with, you would have the officient, you would have two spouses
or more depending on your jurisdiction, and you probably have some
witnesses. So any of those subjects could have their own confidence method.
And so it's really the subject in the RDF sense of it that we're using here.

Denken Chen: Yeah, if you wor you can go ahead

Ted Thibodeau Jr: Yeah, I think where I struggle is that we're saying that
this confident confidence method could be applied to any entity in the RDF
subject sense of the subject. not necessarily the presenter, certainly not
the verifier because they're out of play at that point.

Ted Thibodeau Jr: It's the level of non-conrete that is going on and so
whomever is trying to understand this as a verifier, right? it gets
presented to me and there's a confidence method here somewhere that applies
to something in the credential and this is what I said going back to the
beginning of confidence method is that it seemed to me that it could be
applied to any subject in the RDF sense that is mentioned in the VC that
this could

Ted Thibodeau Jr: confidence in that the entity being identified. Tongue
twister there. I can't lie down on the road on this because I don't have a
firm enough I don't have my feet firmly enough under me to say this is the
problem. All I can say is this continues to reconfuse me. So that seems to
me to be an ongoing issue. Go ahead, man.

Manu Sporny: Yeah, I mean a plus one to that. I mean, like you said, Ted,
you've been involved in this work for a very long time. So, if you are,
struggling to understand what we're meaning with the words we're putting on
the page, that's a problem, right? Because it means that most certainly
other people will be as plus trying to address so that's one. statement too
is what Joe said and Denin and Dave said aligns very much with my internal
mental model of this thing that we're talking about. it really is meant to
be about any subject in the verifiable credential and as we all know a
verifiable credential can have multiple subjects.

Manu Sporny: I think Joe, your example of a marriage license is a great
example because it does have multiple subjects that it talks about. It's
the officient each spouse each witness, right? So those are multiple
subjects and each one of them might have a confidence method associated
with them. that the issuer doesn't necessarily establish when it the issuer
could have established that confidence method in the individual two five
months before, a year a picture, the issuer might have a portrait image on
file for a specific subject and they established that many months before
they actually issued the credential and when they issue the credential they
include the confidence method with it.

Manu Sporny: Plus one to the concept of framing the spec as this applies to
any subject in a credential right the holder can be one of them but that is
a specific case of the more general case which this is meant to be attached
to any subject in any verifiable credential there can be multiple subjects
so maybe what we do we don't use the dog collar use case maybe We use a
marriage license use case and we very clearly point out like these are the
subjects there's the officient they're the spouses they're the witnesses
and each one of them can have a confidence method on them and that's the
concrete thing that we use to ground the text here.

Manu Sporny: And then Ted hopefully that is concrete enough. and then if we
need to make further changes, we make further changes later, Ted. But I
think I'd need to understand from you this is what's wrong with the new
text, So if I update this and I try to be very clear about when we say what
a conference method is associated with it's the subject blah blah blah
basically exactly…

Manu Sporny: what I said if I put that in the PR would that be at least
good enough for now Ted

Ted Thibodeau Jr: I think it's going in the right direction.

Ted Thibodeau Jr: And part of what needs to be clear This is subject in the
VC sense. I think that will help.

Ted Thibodeau Jr: I hope …

Manu Sporny: But it's not.

Manu Sporny: But It really is subject in the RDF sense. So, I'm wondering
why you went there why do you think it's not subject in the RDF sense?

Ted Thibodeau Jr: because subject in the RDF sense could apply to any
entity that is mentioned in any capacity within the VC.

Ted Thibodeau Jr: It's not just one of the subjects of the VC.

Manu Sporny: You're right.

Manu Sporny: But that's true.

Manu Sporny: The confidence method can be attached to any subject mentioned
in the Full stop.

Ted Thibodeau Jr: which is not just subjects of the VC.

Manu Sporny: It's not just correct. I thought you said this is about just
subjects of the VC.

Ted Thibodeau Jr: that the text you wrote in the air that I heard was
talking about the subjects of the VC, right? You had the witnesses, the
participants, and the officients,…

Dave Longley> to give a non "credential subject" example, suppose the VC
has software that was used to issue it -- and a company or developer
mentioned who created that software

Manu Sporny: Mentioned in the graph in the VC.

Ted Thibodeau Jr: and but they are all in the VC sense of subject. it is
not just that they appear in a subject position within the graph of VC
because there can be other things right.

Dave Longley> attached to the VC (we need a better example, but just trying
to talk about one)

Ted Thibodeau Jr: You can have a graph in the BC which has a subject in the
VC sense u attribute entity in the RDF world and that entity can be a
subject of another statement C or rather in the VC graph but it is not a
subject of the VC. Maybe it's too much in the air.

Dave Longley: So, I think we probably need another different example than
just I think the marriage certificate one is very helpful in showing
multiple credential subjects, but that doesn't get at the thing that's
bothering d. I have a bad example that we shouldn't use, but I'm just
bringing it up here so that we can distinguish it because it was the first
one that came to mind. If for some reason in a VC you wanted to attach at
the top level of the VC a claim that says this VC was issued by this
software and this software was created by this developer and here's a
confidence method for that developer.

Dave Longley: You could do all that. anyone can say anything about anything
in a VC. And that would be such both the software and that developer would
both be or subjects in the RDF sense that are not subjects of the
credential. And you could put a competence method there would be a
legitimate use of the term. We need probably a better example than that,
but that would show that you can use it like that. And I don't know why
this would be a prominent use case at all,…

Ted Thibodeau Jr: I think that's not a bad example at all.

Dave Longley: but you could imagine someone needing to assert, I've got
this credential and it was issued by my company and I built the software
and so now I can present that, for example.

Manu Sporny: Yeah, that's a plus one to that. So, I don't know, if that's
helping clear things up. the other approach is, the issuer. We could put
confidence method on the issuer and that's not the credential subject in
the VC, but it is a subject in the RDF sense and I think that would be a
totally legitimate kind of use of the confidence method. What I am
concerned about is showing people very like atypical non-standard things to
make this point Ted right because if we show people and the software gets
its own confidence method and the issuer gets its own confidence method I
don't think we in the group think that that is going to be a very common
use case at all and then I'm conf concerned about that confusing people
where they're just like

Manu Sporny: I'm supposed to do this with my issuance software. and I guess
I as an issuer have to include a confidence method. we're kind of teaching
the wrong thing in the spec at that point. but I don't know, if that helps
clear up when at least I say confidence method attached to a subject,…

Manu Sporny: I don't make the distinction that you're making a subject of
the VC. I wasn't making that distinction, you had

Ted Thibodeau Jr: Yeah, I think that is helpful in going in the right
direction of…

Dave Longley> everyone gets a confidence method!

Ted Thibodeau Jr: what needs to be expressed. I think the software example
again the handwavy parts it becomes possibly not what was originally meant
but I think that if you consider the software example in the Linux kernel
development world when you submit a line of code there's a level of
confidence

Ted Thibodeau Jr: that the recipient needs to know that it's from you
because of the licensing that's involved there and the potential for that
code to actually be put into use managing a nuclear power plant. these
various things and the confidence needs to be there that this line of code
came from that coder because if something's going goes wrong they need to
be able to trace it back and say where's the bug?

Ted Thibodeau Jr: What's wrong with this? Is this what you actually wrote
that went in that kind of thing? I think if this gets fleshed out in words
on screen, that will probably help not just me but others. So, I'll ask
that yes, we go in that direction and…

Ted Thibodeau Jr: see what happens next.

Denken Chen: Okay, thanks.

Denken Chen: So, it's a nice conversation here and yeah, man, please

Manu Sporny: So, I'll take an action to try and update this and I'll try to
put those concepts in there, Ted. And then if we can maybe iterate on it
async, if the group's okay with us kind of like getting to the point where
Ted's like, "Yeah, this is good enough for now." I would like to merge this
sooner than later. I've got a number of other questions around our timeline
and whatever. And I think we need to get at least the things we need to get
the core set of features we're attempting to get into the spec and move
forward on thread model and ask for horizontal review. So that's kind of
what's driving some of my I would like to see this stuff move forward. I
don't want to spend another couple of weeks talking about this. so if
that's okay with folks I will make another pass at this to try and
incorporate language.

Manu Sporny: I think that I'm hearing from you Ted and then if you can kind
of engage as quickly as you can to iterate and the second Ted's like this
is good enough for now I'll merge or if the editors can merge and we can
move forward that would be… what I'd love to see. That's it. Manu Sporny:

Denken Chen: Nice. Okay.

Denken Chen: Thanks. let's move to another probably related things. So for
this PR I'm trying to emphasize that Ted mentioned we originally has a
broader definition of this confidence methods and so I bring it back and I
also trying to make it clear that making sure the current presenter is the
subject is one of the Man,

Denken Chen: use case for our BC competence method but we mentioned
marriage we mentioned at certifications is also a use case so I'm wondering
that this kind of origin is good enough for us to start with a kind of more
gener generic description for this spec and down to specific several use
cases we can elaborate it more in Manuel mentioned you'll be drifting some
wording on it or probably we can add more concrete example for each of the
example there. Yes.

Denken Chen: So yeah,…
Pull Request 45 Discussion Align with the original VCDM 2.0 definition by
denkeni · Pull Request #45 · w3c/vc-confidence-method · GitHub
<https://github.com/w3c/vc-confidence-method/pull/45>

Denken Chen: please that right and…

Ted Thibodeau Jr: as that renders right now.

Ted Thibodeau Jr: I'm fine with it.

Denken Chen: yeah menu and de mer soon.

Manu Sporny: Yeah, I already did a approval on the PR.

Dave Longley: Yeah, I also approved

Denken Chen: Yeah.

Manu Sporny: Yeah, plus one to that. I do have other little nitpicks in the
initial paragraph, but that's a separate PR. I think this is a good

Denken Chen: Yeah. Yeah, it's Very nice. Thanks. Peace.

Ted Thibodeau Jr: And just so people know, there is something wrong with
the Roy respect runs such that I don't get to see the pretty previews nor
even the editor's draft.

Ted Thibodeau Jr: s nor not final publications anything on GitHub IO does
not finish processing through respec for me I get to see the previous
iteration of the code with the triple square brackets instead of the
citation for instance so I don't get to see this until just now and that's
why I can Okay, this right now looks good to me. I haven't looked at the
source to say that's perfect, but that's fine. We'll go from now.

Denken Chen: Yeah, it's running

Ted Thibodeau Jr: That's it.

Manu Sporny: Yeah, just a little side quest, but Ted, c can you not click
on the preview and diff things in the PR? Denin, can you go to where you
got to this page from?

Ted Thibodeau Jr: I can click on them and I will do that now just to be
doubly sure that the problem still exists. so when you click on either one
of those links, you get to a page that starts off with some boilerplate and
the title of the document and goes on Town.

Ted Thibodeau Jr: I see that,…

Ted Thibodeau Jr: but I don't see it styled. The CSS doesn't do the right
thing. And I think that that's because of respec not doing the right thing.
The only error that I find when doing this is there's a cause error
associated with respect and that's the closest thing that I found to an
identifiable problem. But other people have said that they see the thing
render fine with a cause error. God knows.

Manu Sporny: And it shouldn't be running any respec it should be a
completely static page that you're looking at. good data. all these tools
are broken in various different ways and I'm trying to help LE3C fix it,
but it's really hard to track down because we don't have access to any of
the runtime environment or the logs or anything. So, it's a Manu Sporny:

Ted Thibodeau Jr: I hear I just wanted to let folks know that I'm not
intentionally being a problem.

Ted Thibodeau Jr: It's literally that I can't see the thing that you can.

Denken Chen: Okay.

Denken Chen: When I creating the PR, I will look closely about how this
preview and div works correctly. So probably there were times that it will
break. so whenever it happens you can I think different a message there
probably just a new commit will rerun the whole deploying process to help
fix that. Yeah, man. Please.

Manu Sporny: That is unlikely to fix Ted's problem, but yes, you can try
that. Denin, I don't think the issue is with anything you did. Dave did
bring up a good point. Ted, are you running any extensions whatsoever? And
have you tried running looking at it in private browsing mode with all
extensions turned off? Extensions are a very common thing that breaks
everything.

Ted Thibodeau Jr: I don't know an easy way to turn off all the extensions
for a given tab. I have tried it in a private browsing window and that did
not solve the problem.

Manu Sporny: Okay, thanks.

Ted Thibodeau Jr: I have many I'll see…

Manu Sporny: Usually I fire up a browser I don't use at all which has
nothing installed on it and try that.

Ted Thibodeau Jr: if I can find one. Part of my glorious job is testing
things in every browser that I can run.

Ted Thibodeau Jr: So that's

Denken Chen: Okay, thanks.
Biometric Image Section

Denken Chen: Okay, and then we'll move on to biometric image section. I
think there's problems here. so if there is no other issues raised about
this, I will also merge this as well. So this part I just add a simple
biometric samples for this. So we identify the properties there and expand
the sample there.

Denken Chen: So it should be pretty trivial and in this case we are really
limiting to the biometric vitality to face only and we are in intended to
do so because trying to limit the biometric image to phase and then we have
Scott's biometric image vector method that could deal with some other
things to keep people aware that biometric information should as private as
possible. yeah, so that's for this So no problem for So the last one I
think there will be some discussion over this.

Denken Chen: Almighty priest.

Manu Sporny: I think we should hold on one second. I'm trying to make sure
we topic this so it's attached to the PRs. so Denin, when you switch topics
in the agenda, try to put topic colon and then the GitHub issue or the
GitHub PR. Otherwise, our conversation doesn't get attached to the issue.
just I put in chat just now. this one's fine. we can put it in, but one of
the things you said is a little concerning to me. meaning the biometric
image I think could be in the future iris or fingerprint as well because
those are other biometric and maybe I don't know if those are biometric
templates and not biometric images.
Add biometric image section by denkeni · Pull Request #44 ·
w3c/vc-confidence-method · GitHub
<https://github.com/w3c/vc-confidence-method/pull/44>

Manu Sporny: I don't know if you think you feel there's a difference there
so it's fine as a first PR right but I'm want to make sure that we are
crystal clear about are we just talking about pictures or are we talking
about templates as well and maybe a fingerprint is a biometric vector and
not a biometric image and so I'm

Manu Sporny: wondering if other types of biometric images are full is
potentially an iris because that includes color information not just vector
information so thoughts on that Denin I thought we were doing various
different types of images here but you're saying we're only supporting face
…

Manu Sporny: I guess the expectation is that we would add other things like
Iris in the future or thoughts

Denken Chen: Yeah. the processing facial image is a strong request from
various credentials out there you have face image from the passport from
driving license etc things and…

Denken Chen: it's a low hanging fruit for onsite verifications.

Denken Chen: but for iris or several other fingerprint things it's a lot
sensitive information that usually only used by the war coin is using it
the border countries using fingerprints however when I'm researching about
the IO digital travel credentials in their spec they deliberately limiting
putting the biometric information to facial image only.

Ted Thibodeau Jr: This is

Denken Chen: So this aligns with our in a foreseeable future if we are
going to research how a format passport things can put it into a verifiable
credentials the facial image is good enough for this. So I'm wondering if
we got any requests from how could we make iris or fingerprints on the VC
and is it good enough for our spec to protect with just sedative
disclosure? I'm not sure about that.

Ted Thibodeau Jr: Thank you.

Denken Chen: Yeah, please.

Manu Sporny: Plus one to all those concerns. I just don't want us to paint
ourselves into a corner. So if we only think of the face the portrait image
and we're building this around that we are going to paint ourselves into a
corner when inevitably somebody else says I have a different type of
biometric other than face that I want to use. if that happens we don't have
to go through a completely different design. We don't want them creating
another type for this.

Manu Sporny: it'll just lead to interrupt issues it will have to go through
what is it called another standard cycle that sort of thing so I am
concerned about not solving the problem in a generalized way for all
biometric is. So that's Statement two is that I think we need to be very
careful about the language we put in the spec and let people know how
concerning andor dangerous these types of things are, right?

Manu Sporny: Ideally, you're not using any there are some use cases that
require them, but we don't want them to fall into everyday usage. So,
biometric portrait on a passport is and driver's license is what's done
today. But I think we need to be very clear in the security and privacy
consideration sections that these mechanisms can be abused. It may be that
we say if you're going to do biometric image confidence method, you
absolutely must not provide it in a full disclosure credential, you must
use selective disclosure and software must u not provide this by default. I
think we need to really tighten up the language around the usage of some of
these confidence methods.

Manu Sporny: So a cryptographic keybased con confidence method I think
we're largely okay with that image only for super high risk use cases where
you absolutely need to know what the person looks like and you must not
send it over a full disclosure mechanism and…

Denken Chen: Natural Peace. Manu Sporny:

Manu Sporny: and so on and so forth. We'll stop there.

Joe Andrieu: Yeah, I'm a little concerned by the scope that you've just
increased here,…

Joe Andrieu: This was intended specifically just to be about images and it
was intended to be simple and trying to solve this for all possible images
is not something that we have in scope. we do have a biometric modality
here. So there is a point of extensibility such that other items can be
addressed. face is not retina and so I think the use case that Denin got
involved in this work to address is to have a face in an image and I want
that image in the VC and so how do we handle that? That said, I agree with
all your concerns about privacy considerations and letting people know that
you should be careful about using all of this in ways that we don't have to
say quite the same things that we do for the cryptographic mechanisms.

Joe Andrieu: So I do think we need those sorts of considerations in here,
but I don't think we need to define iris based biometric modalities in the
specification. I think that's overkill.

Dave Longley: So I agree that we don't need to go and define a bunch of
these. I think we should just modify the biometric modality language here
to not say that in this version of the specification the value must be face
but instead it should be this specification defines the following
modalities in this table and…

Denken Chen: Yeah.

Dave Longley> can the "biometricModality" section just say "see this table
for values defined by this specification" -- and some language to allow
other values to be created/used in the ecosystem

Ted Thibodeau Jr: Yes. Dave Longley:

Dave Longley> +1 to talking about privacy considerations, etc.

Dave Longley: face can be the only one that's in there. and then make it
clear that that can be a point of extension for people to innovate and do
what they

Manu Sporny: Yes, plus one. Thank you, Dave. That was the point I was
trying to make because the current language means that it's not extensible.
right, Joe?

Joe Andrieu: Yeah, it is bad language. I mean, it's trying to say in future
versions,…

Joe Andrieu: we'll amend it, but that is not quite the same level of
extensibility.

Denken Chen: No far,…

Denken Chen: I will make the adjustment to this sentence and then let the
folks know and reach it later. Thanks for your input. so for the last one,
we have sections and the titles.

Denken Chen: So we originally have the n competence method and then we
rebrand ourself into VC confidence because we are adding assurance levels
and so when we bring in that to VC working group I mentioned it would be a
pain to really need to change the short URL there…

Ted Thibodeau Jr: Sorry. What's the PR here?
PR 47: Title and Assurance Level

Denken Chen: but the short URL is only matters with the URL not with the
titles. So we can pick any title as we want. P is 47 or plish. Thank you
for adding a topic. Yes. So I think The first is a title whether we should
change a title. And the second thing I have to bring it whether that the
assurance level should be at the same level of competence methods or it
should be the child of it.
Editorial cleanup: sections, title, tooling by denkeni · Pull Request #47 ·
w3c/vc-confidence-method · GitHub
<https://github.com/w3c/vc-confidence-method/pull/47>

Denken Chen: And from my understanding the assurance level is usually one
level per is insurance. So it should be tied to one VC not to subject or
confidence methods. So that's why I putting it into this say to look at
this is we have confidence method and then we have assurance level should
be separate and then about title now I'm putting it as confidence and then
we keep the short URL as confidence methods. Is that okay with the group?
Yeah.

Denken Chen: increase.

Manu Sporny: I suggested some changes. so the pattern that we followed in
the render method specification was verifiable credential render methods
plural version 1.0 IO. this specification does define multiple different
methods. the concern I have is it also provides multiple different
assurance levels. So the title might not be right. I'll mention that we
don't have to follow the title in the charter and in fact often or on a
regular basis groups sometimes rename the specifications as they work on
them to get more accurate titles. so I don't know if anyway interested in…

Denken Chen: That's

Manu Sporny: how the title resonates with the rest of the

Dave Longley: One of the concerns that was raised by the working group just
briefly by Phil Archer was that just calling this verifiable credential
confidence is too broad and I think I agree with that. there are many ways
in which you can apply the term confidence and you could think of the
recognized entities specification as something that helps establish
confidence in whatever you're accepting in some sense. And so I think we're
going a little too broad if we do that. And I also don't think it's a
problem for us to put confidence methods in the title of what we're doing.
And I don't think it's a problem that we say that, in this specification,
we can use other elements of other things, evidence, whatever it is we want
to do to accomplish things. I don't think that's a problem and…

Denken Chen: Yeah, sure, please.

Dave Longley: just because the title has confidence methods in

Joe Andrieu: Yeah,… I think If we were to get rid of confidence methods or
if we were to put confidence methods in then I would like to also put
assurance levels in confidence was meant to address all of those since we
don't define just confidence methods. So I think it is overly constraining
to have the methods in there without pointing out we're also defining these
other things. Joe Andrieu:

Dave Longley: If we say things in evidence, I'm not sure how we're going to
model any of this yet. I worry about our title becoming a list of all the
things that we currently say in this spec. so I think just using confidence
is ly I think having to enumerate everything in the title that we're then
going to put in the spec is not a good use of a title. And I don't know
what the middle ground is there. But I felt like we're primarily talking
about confidence methods and any other thing that we do is at least in some
sense related to that. but I'm not going to die on this hill if we make a
big long title that puts a comma separated list of everything we're
defining. But I don't think that's a necessarily good thing.

Dave Longley: And I can see that list growing that if we make some new
thing that's called assurance level we've got confidence method What…

Denken Chen: Next show, please

Dave Longley: if we need to model something else and we add that thing? Now
we've got to update the title.

Joe Andrieu: Yeah, I agree that the laundry list is problematic,…

Joe Andrieu: but that's why we went with just confidence. I mean if we want
to do the work to come up with the aggregate term I could support that but
deprecating assurance levels such that is not what this is clearly talking
about makes me feel like a preamble to cutting that out of the spec
entirely and I would oppose that.

Ted Thibodeau Jr> +1 "Method" (or "Methods") should be kept in the tile.

Joe Andrieu: So I think we still have work to do to come up with what's the
aggregate term that really this document is saying here's how you can have
some reliance on this data in different ways either reliance on a secondary
method which is a confidence method that you're going to execute and trust
or reliance on something the issuer is attesting to and…

Manu Sporny> Establish Confidence In Subjects :P

Denken Chen: Man, please

Joe Andrieu: so you feel more comfortable relying on their attestations but
I don't know that that's a direction to help either. That's all.

Manu Sporny: Yeah, I mean plus one on that. I don't think anyone's arguing
to remove the assurance stuff from the spectra. So I would also oppose
that. I think it's good to have it in there. I'm, of course doing, AI based
searches on trying to find words between that mean confidence and evidence
at the same time. And it's, cogent, ta corroborated, justified,
incontrovertible. none of them useful. So, we're in the middle of mic
shedding exercise. Maybe we can just pause this and, think and come back to
it and spread this out over multiple weeks. that's it.

Dave Longley: Just to throw one more thing in there is maybe we're broadly
talking about subject authentication in the specification.

Dave Longley> maybe this is just "Subject Authentication"

Denken Chen: I think I'm good with either way and I'm more inclined to add
the methods back and see where we could go. And so like I mentioned the
assurance table is usually issued per issuance not per subjects usually. So
I'll keep it as I know that the overall concept is probably little bit
messy there but we all agree that a competence method is not really useful
without assurance level informations.

Ted Thibodeau Jr> *`property of a "Confidence Method"`, among other things

Denken Chen: So for example, we push a little bit on visitor car that is
actually one and that's pretty weak. but it's good enough for visitor
management systems. But when it comes to driving license, the insurance
level is important message to the verifier. So hopefully we got consensus
over there. it's very productive. so thanks all for coming. do we have any
other things to raise? Yeah, man. Please.

Manu Sporny: Yeah, I want to know what the timeline for horizontal review
So, if we get these things in there, the only thing that we don't
necessarily have is a threat model. I can take an action to do that or I
think maybe Joe you were going to do that. I can't remember. and in the
assurance level bits feel like it's kind of a gap. We need to get that a
little further along. I don't know if it's far enough along for horizontal
review yet. So, if we're multiple weeks out from horizontal review, I kind
of want to understand what are the things we're going to be focusing on to
get us to horizontal review before TAC. That's it.

Joe Andrieu> Subject Authentication is pretty good

Manu Sporny> "Authenticating Subjects in Verifiable Credentials v1.0"

Denken Chen: Yeah, I would like to mention that and to merge these four PR
within a week with some little modification there and with the biometric
image and the cryptography keys

Denken Chen: if it's our current scope and so we should be ready for third
model and Joe would you like to briefly talk about your strategy It's true.

Joe Andrieu: …

Joe Andrieu: I want to talk about something else. I didn't track what you
just queued me up for. I think I want to help, the two editors, we've been
very consistent that we'll be ready for horizontal review at TAC, which is
10 weeks from now. And I think we can get I do think we should enumerate
what are the things that we need to do to get there. and the threat model
is one of them. but you opened with the question of if you want to know
what our timing was about horizontal review and the timing is we will be in
horizontal review by TAC.

Ted Thibodeau Jr: Let's go.

Joe Andrieu: Okay. thumbs up on that. And then I guess Dankin,…

Joe Andrieu: we should take that on for our editor's call just to put
together a punch list of what do we need to do to get into HR. because that
would be useful. Cheers.

Denken Chen: Yeah, no problem.

Denken Chen: Okay, thanks for coming and see you in two weeks. Meeting
ended after 01:00:23 👋 This editable transcript was computer generated and
might contain errors. People can also change the text after it was created.

Dave Longley> would be fine with that ^

Manu Sporny> (would be okay with VC Subject Authentication v1.0 as well)

Dave Longley> yup, also fine.

Ted Thibodeau Jr> that can easily be read as versioning VC, rather than the
itended AS (or ASinVC).

Ted Thibodeau Jr> Complex version levels are hard to address in titling.

Dave Longley> agreed ^

This transcription was generated by a large language model (LLM) and might
contain errors. When in doubt, check the audio recording. This page was
formatted by scribe.perl <https://w3c.github.io/scribe2/scribedoc.html>
version 248 (Mon Oct 27 20:04:16 2025 UTC).