[SURBL-Discuss] Pleaae beta test ds.surbl.org - 6dos data

Jeff Chan jeffc at surbl.org
Tue Jun 29 08:23:10 CEST 2004


On Tuesday, June 29, 2004, 7:12:38 AM, Chris Santerre wrote:

>>From: Jeff Chan [mailto:jeffc at surbl.org]

>>As a data point, 6dos hit 300 whitelist entries out of 120,000
>>records, which is about a ten times greater whitelist hit *rate*
>>than ob.surbl.org.
>>

> 0.25% fp rate, so it has an S/O rating of 99.75 :) 

No, that's not an FP rate since my whitelist does not include
every possible FP.  In fact, it's rather limited.  More like
the 1000 most common web domains plus many more obscure
geographic tlds that will probably never be used in spams.

The whitelist hits might give a hint at relative FP rates between
lists, but only actual testing against real messages will give
meaningful FP rates.

> Actually that is great info. Can we get the whitelist hits? This might be a
> great way to tweak the 6dos list. I'm also very interested in who hit the
> whitelist. I'd like to see the xref in 6dos to see who these people are
> dealing with. I think RSK would be interested as well.

I've saved a copy of the 6dos hits against my whitelist at:

  http://spamcheck.freeapp.net/6dos.domains.whitelist-hits

The entire whitelist, including many geographic domains is at:

  http://spamcheck.freeapp.net/whitelist-domains.sort

> Even if we have to clean up 1-2% of these listed, look how many evil domains
> we get. But I fully understand your philosophy on this Jeff. Some of these
> evil domains may not have spammed.....yet. ;)

I don't mind pre-emptively listing every domain of every known
spam operation.  What we don't want are FPs on legitimate domains.

Jeff C.



More information about the Discuss mailing list