[SURBL-Discuss] Testing: xs.surbl.org (CBL Data)

Jeff Chan jeffc at surbl.org
Tue Apr 19 16:31:42 CEST 2005


On Tuesday, April 19, 2005, 6:43:56 AM, Alex Broens wrote:
> Guys,

> Jeff's little XtraSmall (CBL Data) lust, ehhhmm, list seems to be 
> catching quite a bit of trash.

> give it a try:

> urirhsbl URIBL_XS_SURBL xs.surbl.org.  A   2
> body     URIBL_XS_SURBL eval:check_uridnsbl('URIBL_XS_SURBL')
> describe URIBL_XS_SURBL URL listed in XS SURBL - TEsting
> tflags   URIBL_XS_SURBL net
> score    URIBL_XS_SURBL 1.5

> Score to your taste !!!!!

> Pretty sure Jeff is anxious to get FP reports.

> h2h

> Alex

Thanks much Alex!

BTW, the list is about 1k records at the current levels, and it
may be a little misleading to talk about 100 new records, because
the total records in each category are greater:

Without any processing the current list has about 9k records:

    8781    8781  111633 cbl-domains.all

Taking the 97th percentile of volume-ranked hits gives:

     565     565    8416 cbl-domains.percentiled

The intersection of all with existing SURBLs is:

     906     906   13459 cbl-domains.surbl

And the intersection of the percentiled and SURBL hits is

     991     991   14817 cbl-domains.afterwhitelist

1k from 9k may seem like we're losing a lot, but the
distributions look Zipfian:  A few records get many hits and many
records get a few hits, so there's a lot of "noise" down in the
"few hits" range which may not be very useable.  And even at this
conservative setting, we're getting 97 percent of the CBL URI
trap hits by volume, which can't be too bad.

Cheers,

Jeff C.
--
"If it appears in hams, then don't list it."



More information about the Discuss mailing list