We can probably significantly reduce the false positives on ob.surbl.org, the SURBL list based on Outblaze's URI blacklist:
http://www.surbl.org/lists.html#ob
at the cost of some possibly minor false negatives.
SpamAssassin has live (weekly?) statistics about the performance of their rules, including all SURBL lists, against their ham and spam corpora at their Rule QA site:
http://ruleqa.spamassassin.org/
As you can see OB ranks significantly below the other SURBL lists, with much higher FP rates around 0.1% compared to 0.01% to 0.025% or so. (Note that the Rule QA site seems to have occasional glitches, so if the numbers seem out of range one week, check again later.)
Should we make that change?
Jeff C.
On 12/07/2008 4:41 AM, Jeff Chan wrote:
We can probably significantly reduce the false positives on ob.surbl.org, the SURBL list based on Outblaze's URI blacklist:
http://www.surbl.org/lists.html#ob
at the cost of some possibly minor false negatives.
SpamAssassin has live (weekly?) statistics about the performance of their rules, including all SURBL lists, against their ham and spam corpora at their Rule QA site:
http://ruleqa.spamassassin.org/
As you can see OB ranks significantly below the other SURBL lists, with much higher FP rates around 0.1% compared to 0.01% to 0.025% or so. (Note that the Rule QA site seems to have occasional glitches, so if the numbers seem out of range one week, check again later.)
I'm not aware of any glitches. Is it possible that you're seeing the numbers change based on who or how many people submitted results that week?
Should we make that change?
You didn't really describe the change, so, sure. :)
Daryl
--On Saturday, July 12, 2008 1:41 AM -0700 Jeff Chan jeffc@surbl.org wrote:
We can probably significantly reduce the false positives on ob.surbl.org, the SURBL list based on Outblaze's URI blacklist:
http://www.surbl.org/lists.html#ob
at the cost of some possibly minor false negatives.
SpamAssassin has live (weekly?) statistics about the performance of their rules, including all SURBL lists, against their ham and spam corpora at their Rule QA site:
http://ruleqa.spamassassin.org/
As you can see OB ranks significantly below the other SURBL lists, with much higher FP rates around 0.1% compared to 0.01% to 0.025% or so. (Note that the Rule QA site seems to have occasional glitches, so if the numbers seem out of range one week, check again later.)
Should we make that change?
This agrees with our experience using SURBL for a few years. We've seen the occasional fp with the ob listings, and none I am aware of with the rest of SURBL.
What's the change?
Joseph Brennan Lead Email Systems Engineer Columbia University Information Technology
-
Daryl C. W. O'Shea -
Jeff Chan -
Joseph Brennan