[Asrg] REDIRECt to ASRG RE: Reputation systems

["Hallam-Baker, Phillip" <pbaker at verisign.com>]

I think that this is getting into an ASRG question.

If we create a graph of false positives vs false negatives we get
an L shaped curve for most filtering technologies that are
parameterizable and a point for a non perameterized technology
such as a blacklist.

Rather than arguing this point round and round does anyone have
a ready made graph with the various blacklists plotted on it so
that we could make a useful comparison?

There are two reasons why a list would have more false positives,
the listing criteria/application thereof could be wrong or the
list could be more aggressive in listing more spam sources.

If List-A listed 100 addresses and 80 were wrong this would result
in fewer false positives than List-B listing 1,000,000 addresses
with 800 wrong. So there is a discrimination function here...


> >We got a lot of false positives with the SpamCop list, and other BLs
> >are bound to be worse.
> 
> Actually, Spamcop's list is infamous for its vast number of false
> positives.  That's why nobody I know uses it for anything more than
> a fairly low-weight input into a scoring system.
> 
> Try the Spamhaus SBL and XBL and you'll find very few misfires.



_______________________________________________
Asrg mailing list
Asrg at ietf.org
https://www1.ietf.org/mailman/listinfo/asrg