Re: [stir] Adding categories from sipcore-callinfo-spam to draft-ietf-sipcore-callinfo-rcd-04

[Henning Schulzrinne <[e-mail]>]

The basic idea originated in the FCC robocall strike force a few years ago,
with lots of carrier participation, so you could go back to the various FCC
reports that were produced at the time for the motivation and background
(e.g., https://transition.fcc.gov/cgb/Robocall-Strike-Force-Final-Report.pdf
and https://www.fcc.gov/file/12311/download). This I-D and requirements
have been discussed since 2017, so this is hardly a new idea. The draft
itself has been through IESG review - I just never got around to implement
the changes requested.

The basic requirement is that we want the receiving callee to make
automated or semi-automated judgements. Currently, the RCD draft has a
free-form field, but that doesn't really help for anything automated (or,
say, displaying icons or other graphical elements).

Like all spam labeling, this is probabilistic, i.e.,  if you're waiting for
perfection, this isn't for you. Current robocall filtering services already
label numbers with categories (they are in the same ballpark, but not
exactly the same and each seems to differ a bit). There are at least two
possible sources:

* For some labels, the calling party (with signing) will insert the
information, presumably for the more "positive" labels like "health" or
"government" or "personal". If a carrier lies, this becomes either a
reputational issue ("never trust labels from carrier X") or enforcement
matter (e.g., as a potential deficiency in robocall mitigation). The label
may be part of the KYC process when a carrier signs up a customer - after
all, in almost all cases this is pretty obvious ("your business says Joe's
Travel Agency. You are picking the healthcare designation how?") A
terminating carrier may, for example, decide to only trust known carriers
(say, T-Mobile) in deciding whether to convey this information to the
called party.

* An intermediary service used by the terminating carrier labels (signed)
numbers based on honeypots, crowdsourcing or number databases, probably
more focused on negative labels like "telemarketing" or "fraud". Again,
versions of this exist today.

Henning

On Sat, Mar 12, 2022 at 3:10 PM Holmes, David <[e-mail]>
wrote:

> Hi Henning,
>
> Interesting idea, but who could define the categories, and who would
> verify the claims?
>
> There may be something here, but this is a good example of where we should
> agree the use cases and then define requirements before leaping to
> solutions.
>
> BR/David Holmes/T-Mobile USA
>
> Get Outlook for Android <https://aka.ms/ghei36>
> ------------------------------
> *From:* stir <[e-mail]> on behalf of Henning Schulzrinne <
> [e-mail]>
> *Sent:* Saturday, March 12, 2022 9:24:47 AM
> *To:* [e-mail] <[e-mail]>; SIPCORE <[e-mail]>
> *Subject:* [stir] Adding categories from sipcore-callinfo-spam to
> draft-ietf-sipcore-callinfo-rcd-04
>
> [External]
>
> In trying to get back to looking at my ancient, dusty draft, a thought: I
> think RCD would be significantly more useful if it contained standardized
> categories of callers, such as the one included in callinfo-spam. This
> allows much better user interfaces and automated handling ("send all
> political calls to voicemail", "play special ringtone for personal calls").
>
> Henning
>
>