Chuck, Here some numbers from my side:
100k messages in the last 7 days 50.5% identified as legit, 49.5% as spam (viruses was filtered out before) The best IP4R-based tests was CBL (21%, 0.37%FP), SPAMCOP (21%, 0.47%FP) and XBL-DYNA (19%, 0.27%FP) So they catch less then 50% of incoming spam without creating a significant number of false positives. FIVETEN-SRC was able to catch 24% of spam but has also had FP's on around 6% of all processed messages. A text-filter combining the results of different IP4R-based tests has reached a catch rate of 36%. I consider it the current maximum that can be reached with IP4r-based tests by having a - let's say - moderate number of false positives. INV-URIBL instead can catch 37% of all messages as spam and I must say that up to now I haven't had time to try improving the INV-URIBL configfile. (Any suggestion is welcome!) It's also important that the number of FP's for this test is near to zero. SNIFFER was able to catch 47% of all spam messages but I must also say that there was a significant number of false positives (5%). Most of them generated by SNIFFER-GENERAL and SNIFFER-RICH. SPAMCHK has had correct results on around 45% of all messages, but also had around 7% of FP's Other excelent tests was CMDSPACE (30%, 1%FP) and HELOISIP (13%, 0.17%FP) Due to Decludes weighting system and the combination of all this tests I can see between 10 and 20 spam messages each month in my inbox, by catching more then 300 spams each day. Markus > -----Original Message----- > From: [EMAIL PROTECTED] > [mailto:[EMAIL PROTECTED] On Behalf Of Chuck Schick > Sent: Tuesday, July 26, 2005 7:57 PM > To: Declude. JunkMail > Subject: [Declude.JunkMail] RBL's becoming worthless... > > In the last several months we have seen large quantity of > spam coming from IP blocks that never seem to get listed on > any RBL. Spamcop is about the only one that picks some of > them up and once in awhile spamhaus. There was a block last > night that sent several hundred and sendbase.org showed they > had detected no email from that block. > > The reason I bring this up is because when we first started > blocking spam I would say the blacklists would catch almost > 90% so we relied heavily on the blacklist. With the > blacklists not being as effective we need to rely on other > tests like sniffer but that misses alot also. > > Chuck Schick > Warp 8, Inc. > (303)-421-5140 > www.warp8.com > > --- > This E-mail came from the Declude.JunkMail mailing list. To > unsubscribe, just send an E-mail to [EMAIL PROTECTED], and > type "unsubscribe Declude.JunkMail". The archives can be > found at http://www.mail-archive.com. > --- This E-mail came from the Declude.JunkMail mailing list. To unsubscribe, just send an E-mail to [EMAIL PROTECTED], and type "unsubscribe Declude.JunkMail". The archives can be found at http://www.mail-archive.com.
