ASRG Identity, Auithentication, and Reputation subgroup

Reputation assignment

In addition to the discussion on lookup conventions there is the problem 
of determining the mechanism for how a reputation service is to compute 
a reputation.  What are the inputs and mechanisms for acquiring and 
assessing those inputs that would allow a service to quantify 
reputation?  What is the formula for doing so?

It seems that one could code a server using SIQ and an SQL database 
fairly easily - but by what mechanism is the reputation score in that 
database supposed to be derived?  This seems to me to be a very 
difficult problem.

--

-- 
Arvel Hathcock
CEO, Alt-N Technologies
http://www.altn.com

-----------------------------------------------------------------
MDaemon 9.0 is coming!  Superior multi-threaded IMAP, AntiVirus,
and AntiSpam performance, Active Directory monitoring, integrated
free/busy server, integrated SyncML server, plus so much more.
-----------------------------------------------------------------

headers

John Levine | 20 Jan 04:23

Re: Reputation assignment

>In addition to the discussion on lookup conventions there is the problem 
>of determining the mechanism for how a reputation service is to compute 
>a reputation.  What are the inputs and mechanisms for acquiring and 
>assessing those inputs that would allow a service to quantify 
>reputation?  What is the formula for doing so?

That's the secret sauce that distinguishes one reputation service from
another.  It seems to me that we're much more likely to make progress
standardizing interfaces to reputation systems than internal
algorithms.

R's,
John

headers

Dave Crocker | 21 Jan 16:57

Re: Reputation assignment

>>     What are the inputs and mechanisms for acquiring and 
>> assessing those inputs that would allow a service to quantify 
>> reputation?  What is the formula for doing so?
> 
> That's the secret sauce that distinguishes one reputation service from
> another.  It seems to me that we're much more likely to make progress
> standardizing interfaces to reputation systems than internal
> algorithms.

Strong agreement.  The challenge is to specify a reporting format that cover the 
likely output of a wide range of reputation services, WITHOUT having to deal 
with the means by which they produce that output.

d/

--

-- 

Dave Crocker
Brandenburg InternetWorking
<http://bbiw.net>

headers

Anthony Howe | 20 Jan 06:27

Re: Reputation assignment

John Levine wrote:
>> In addition to the discussion on lookup conventions there is the problem 
>> of determining the mechanism for how a reputation service is to compute 
>> a reputation.  What are the inputs and mechanisms for acquiring and 
>> assessing those inputs that would allow a service to quantify 
>> reputation?  What is the formula for doing so?
> 
> That's the secret sauce that distinguishes one reputation service from
> another.  It seems to me that we're much more likely to make progress
> standardizing interfaces to reputation systems than internal
> algorithms.

Hai so desu.

SIQ steered clear of server implementation issues for this reason. 
Everyone will have their own idea how reputation should be determined 
from "pay me to say nice things" to "collect data, crunch, spit result". 
  A reputation service will itself develop its own reputation for 
accuracy and fairness over time through market forces and proof of 
method used.

--

-- 
Anthony C Howe                                 +33 6 11 89 73 78
http://www.snert.com/       ICQ: 7116561         AIM: Sir Wumpus

Sendmail Anti-Spam Solutions           http://www.snertsoft.com/
                                             We Serve Your Server