IETF-85 Proceedings
Introduction | Area, Working Goup & BoF Reports | Plenaries | Training | Internet Research Task Force
Additional information is available at tools.ietf.org/wg/repute
Chair(s):Applications Area Area Director(s):Applications Area Advisor |
No Slides Present
In the open Internet, making a meaningful choice about the handling
of content requires an assessment of its safety or "trustworthiness".
This can be based on a trust metric for the owner (identity) of an
identifier associated with the content, to distinguish (likely)
good actors from bad actors. The generic term for such information
is "reputation". This working group will develop mechanisms for
reputation reporting by independent services. One mechanism will be
for a basic assessment of trustworthiness. Another will provide a
range of attribute/value data that is used as input to such an
assessment. Each service determines the attributes it reports.
Various mechanisms have been developed for associating a verified
identifier with email content, such as with SPF (RFC4408) and DKIM
(RFC4871). An existing reputation query mechanism is
Vouch-by-Reference (RFC5518). It provides a simple Boolean
response concerning a domain name used for email. The current working
group effort will expand upon this, to support additional
applications -- such as Web pages and hosts -- and a wider range of
reporting information.
Given the recent adoption of domain name verification for email,
by SPF and DKIM, the most obvious initial use case for reputation is
for email. Inbound email filters that perform message authentication
can obtain a verified domain name and then consult a reputation
service
provider to make a determination (perhaps also based on other
factors) of whether or not the content is desirable and take
appropriate action with respect to delivery, routing or rejection.
Another possible use case is identity-based evaluation of web
content using technologies such as the DKIM-derived DOSETA
(work in progress).
This working group will produce specifications (targeting the
standards track, though the working group will determine the
appropriate status) for:
* the detailed requirements for reporting
* an end-to-end system architecture in which reporting occurs
* the mechanisms and formats for reporting
Two mechanisms are under consideration:
* simple -- a reputation is expressed in a simple manner,
via records in the DNS
(see draft-kucherawy-reputation-query-dns)
* extended -- a response can contain more complex information
useful to an assessor, reported over HTTP using
an encoding such as XML or JSON
(see draft-kucherawy-reputation-query-http)
The syntactic and semantic aspects of mechanisms and formats will be
designed to be application-independent and portable (i.e., reputation
provider-independent). By distinguishing reporting information
(format) from reporting mechanism (channel), the specifications
will permit adaptation to support reporting through additional
channels. Limited application-specific tailoring will be
provided for email, to demonstrate the approach, which can be
applied for supporting additional applications. The design and
specification will also permit adaptation to support reporting
through additional transport channels.
Items that are specifically out of scope for this work:
* Specific actions to be taken in response to a reputation reply.
It is up to assessors (i.e., the consumers of reputation data)
to determine this. Non-normative illustrations, however, can
be included to demonstrate possible uses of reputation data
in a particular context.
* Selection of what data might be valid as the subject of a
reputation query. It is up to reputation service providers and
assessors to select which qualities of a body of data might
be useful input to reputation evaluation.
* Concerns about methods of verifying domain names that are used
for email reputation. A verified domain name is a starting point
for this work; the means by which it was obtained and the
"meaning" of the name or its verification are matters for
discussion elsewhere.
* Algorithms to be applied to aggregated feedback in order to
compute reputations. These are part of a back-end system,
usually
proprietary, and not appropriate for specification as part of
a query/reply framework and protocol.
The initial draft set:
draft-kucherawy-reputation-model
draft-kucherawy-reputation-media-type
draft-kucherawy-reputation-query-http
draft-kucherawy-reputation-query-dns
draft-kucherawy-reputation-query-udp
draft-kucherawy-reputation-vocab-identity