It's sounding like the sc2 list is catching 10-15% more spam than the sc list, based on some early reports of SA users. Is anyone else getting some results?
Are there differences in ham hits?
Has anyone been able to run them through their test corpora?
How about xs.surbl.org?
Jeff C. -- Don't harm innocent bystanders.
On Friday 29 July 2005 10:13 pm, Jeff Chan wrote:> It's sounding like the sc2 list is catching 10-15% more spam than> the sc list, based on some early reports of SA users. Is anyone> else getting some results?>> Are there differences in ham hits?>Here are last nights stats, SC2 is hitting better than SC, no hits whatsoever on ham: Email: 148 Autolearn: 0 AvgScore: 19.63 AvgScanTime: 6.54 secSpam: 103 Autolearn: 0 AvgScore: 32.95 AvgScanTime: 6.39 secHam: 45 Autolearn: 0 AvgScore: -10.87 AvgScanTime: 6.88 sec Time Spent Running SA: 0.27 hoursTime Spent Processing Spam: 0.18 hoursTime Spent Processing Ham: 0.09 hours TOP SPAM RULES FIRED------------------------------------------------------------RANK RULE NAME COUNT %OFRULES %OFMAIL %OFSPAM %OFHAM------------------------------------------------------------ 1 PYZOR_CHECK 103 5.70 69.59 100.00 100.00 2 DIGEST_MULTIPLE 93 5.15 62.84 90.29 4.44 3 RAZOR2_CF_RANGE_51_100 89 4.93 60.14 86.41 0.00 4 RAZOR2_CHECK 88 4.87 59.46 85.44 0.00 5 BAYES_99 87 4.81 58.78 84.47 0.00 6 DCC_CHECK 61 3.38 41.22 59.22 4.44 7 URIBL_SC2_SURBL 59 3.27 39.86 57.28 0.00 8 URIBL_JP_SURBL 57 3.15 38.51 55.34 0.00 9 URIBL_SC_SURBL 56 3.10 37.84 54.37 0.00 10 URIBL_OB_SURBL 55 3.04 37.16 53.40 0.00 11 URIBL_AB_SURBL 53 2.93 35.81 51.46 0.00 12 HTML_MESSAGE 53 2.93 35.81 51.46 2.22 13 URIBL_XS_SURBL 51 2.82 34.46 49.51 0.00 14 RCVD_IN_XBL 45 2.49 30.41 43.69 0.00 15 URIBL_WS_SURBL 40 2.21 27.03 38.83 0.00 16 RCVD_IN_BL_SPAMCOP_NET 40 2.21 27.03 38.83 0.00 17 URIBL_SBL 34 1.88 22.97 33.01 0.00 18 RCVD_IN_DSBL 31 1.72 20.95 30.10 0.00 19 RCVD_IN_SORBS_DUL 26 1.44 17.57 25.24 0.00 20 MIME_HTML_ONLY 26 1.44 17.57 25.24 2.22------------------------------------------------------------ TOP HAM RULES FIRED------------------------------------------------------------RANK RULE NAME COUNT %OFRULES %OFMAIL %OFSPAM %OFHAM------------------------------------------------------------ 1 PYZOR_CHECK 45 25.14 30.41 100.00 100.00 2 BAYES_00 39 21.79 26.35 0.97 86.67 3 AWL 29 16.20 19.59 2.91 64.44 4 SPF_HELO_PASS 11 6.15 7.43 0.97 24.44 5 SPF_PASS 6 3.35 4.05 0.00 13.33 6 USER_IN_WHITELIST 5 2.79 3.38 0.00 11.11 7 RCVD_BY_IP 4 2.23 2.70 7.77 8.89 8 NO_REAL_NAME 4 2.23 2.70 10.68 8.89 9 BAYES_50 4 2.23 2.70 5.83 8.89 10 SPF_SOFTFAIL 3 1.68 2.03 15.53 6.67 11 DNS_FROM_RFC_POST 3 1.68 2.03 17.48 6.67 12 DCC_CHECK 2 1.12 1.35 59.22 4.44 13 RCVD_IN_BSP_TRUSTED 2 1.12 1.35 0.00 4.44 14 DIGEST_MULTIPLE 2 1.12 1.35 90.29 4.44 15 RCVD_IN_BSP_OTHER 2 1.12 1.35 0.00 4.44 16 DNS_FROM_RFC_WHOIS 2 1.12 1.35 6.80 4.44 17 USER_IN_DEF_WHITELIST 2 1.12 1.35 0.00 4.44 18 MIME_HTML_ONLY 1 0.56 0.68 25.24 2.22 19 PLING_QUERY 1 0.56 0.68 0.97 2.22 20 HTML_MESSAGE 1 0.56 0.68 51.46 2.22------------------------------------------------------------ -- ChrisRegistered Linux User 283774 http://counter.li.org09:06:51 up 6 days, 10:08, 2 users, load average: 0.11, 0.09, 0.09Mandriva Linux 10.1 Official, kernel 2.6.8.1-12mdk~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~Texas law forbids anyone to have a pair of pliers in his possession.~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Jeff Chan wrote:
It's sounding like the sc2 list is catching 10-15% more spam than the sc list, based on some early reports of SA users. Is anyone else getting some results?
Are there differences in ham hits?
How about xs.surbl.org?
Since 7/27 config change to add sc2 my server has tagged 27,609 messages as spam: URIBL_SC2_SURBL has hit on 15,981 messages, all of which were tagged spam URIBL_SC_SURBL has hit on 14,105 messages, all of which were tagged spam URIBL_XS_SURBL has hit on 13,165 messages, 3 of which were not tagged spam; however, all 3 look like a fn or damaged virus.
So SC2 catches about 13% more messages for me. In the week prior to 7/25 (when you told me the XS config changed) XS hit on average 1570 messages per day. Since then it has averaged 2,175 messages per day.
On Monday, August 1, 2005, 11:37:19 PM, Daniel Kleinsinger wrote:
Jeff Chan wrote:
It's sounding like the sc2 list is catching 10-15% more spam than the sc list, based on some early reports of SA users. Is anyone else getting some results?
Are there differences in ham hits?
How about xs.surbl.org?
Since 7/27 config change to add sc2 my server has tagged 27,609 messages as spam: URIBL_SC2_SURBL has hit on 15,981 messages, all of which were tagged spam URIBL_SC_SURBL has hit on 14,105 messages, all of which were tagged spam URIBL_XS_SURBL has hit on 13,165 messages, 3 of which were not tagged spam; however, all 3 look like a fn or damaged virus.
So SC2 catches about 13% more messages for me. In the week prior to 7/25 (when you told me the XS config changed) XS hit on average 1570 messages per day. Since then it has averaged 2,175 messages per day.
Thanks for the feedback Daniel. Sounds like SC2 may be worth making into production SC. But I'd still like to hear back from some of the SA corpus testers about ham hit rates.
Jeff C. -- Don't harm innocent bystanders.
Jeff Chan wrote:
On Monday, August 1, 2005, 11:37:19 PM, Daniel Kleinsinger wrote:
Jeff Chan wrote:
It's sounding like the sc2 list is catching 10-15% more spam than the sc list, based on some early reports of SA users. Is anyone else getting some results?
Are there differences in ham hits?
How about xs.surbl.org?
Since 7/27 config change to add sc2 my server has tagged 27,609 messages as spam: URIBL_SC2_SURBL has hit on 15,981 messages, all of which were tagged spam URIBL_SC_SURBL has hit on 14,105 messages, all of which were tagged spam URIBL_XS_SURBL has hit on 13,165 messages, 3 of which were not tagged spam; however, all 3 look like a fn or damaged virus.
So SC2 catches about 13% more messages for me. In the week prior to 7/25 (when you told me the XS config changed) XS hit on average 1570 messages per day. Since then it has averaged 2,175 messages per day.
Thanks for the feedback Daniel. Sounds like SC2 may be worth making into production SC. But I'd still like to hear back from some of the SA corpus testers about ham hit rates.
Jeff C.
Jeff, Have you got some details on the differences you can share with us?
C.
On Wednesday, August 3, 2005, 11:17:43 AM, Chris Chris wrote:
Jeff Chan wrote:
Thanks for the feedback Daniel. Sounds like SC2 may be worth making into production SC. But I'd still like to hear back from some of the SA corpus testers about ham hit rates.
Jeff C.
Jeff, Have you got some details on the differences you can share with us?
Hi Chris, I've not heard back from any SA corpus testers yet. Has anyone else?
Does anyone else have any results for SC2 to share? It seems to hit about 10-15% more spam for different folks, but is it any better or worse on hitting ham (false positives)?
Jeff C. -- Don't harm innocent bystanders.
On Wednesday 03 August 2005 04:08 pm, Jeff Chan wrote:
Does anyone else have any results for SC2 to share? It seems to hit about 10-15% more spam for different folks, but is it any better or worse on hitting ham (false positives)?
Jeff C.
Jeff, from 03:59, 29 Jul to 03:59 3 Aug, I've gotten 901 total messages, 219 being ham which neither sc2 nor xs has hit on. On spam sc2 is between #7 and #10 on spam hits according to the stat script I run. SC only shows up on about 3 of the 6 days. I've extended the hit list to the top 50 which may give better stats.
HTH
Jeff,
Last night, I ran SC2 against a corpus of 119,000 messages consisting of about 97-99% ham. This was a corpus of mostly ham mail which had made it past my filter over the past few months. (BTW, my percentages keep getting better, so most of that small percentage of spam in this corpus was from a few months ago).
SC2 caught about 150 messages from this corpus. ALL of these caught messages were spam. It did NOT "block" a single ham.
--Rob McEwen
-----Original Message----- From: discuss-bounces@lists.surbl.org [mailto:discuss-bounces@lists.surbl.org] On Behalf Of Jeff Chan Sent: Wednesday, August 03, 2005 5:09 PM To: SURBL Discussion list Subject: Re: [SURBL-Discuss] Re: SURBL: sc2 vs sc?
On Wednesday, August 3, 2005, 11:17:43 AM, Chris Chris wrote:
Jeff Chan wrote:
Thanks for the feedback Daniel. Sounds like SC2 may be worth making into production SC. But I'd still like to hear back from some of the SA corpus testers about ham hit rates.
Jeff C.
Jeff, Have you got some details on the differences you can share with us?
Hi Chris, I've not heard back from any SA corpus testers yet. Has anyone else?
Does anyone else have any results for SC2 to share? It seems to hit about 10-15% more spam for different folks, but is it any better or worse on hitting ham (false positives)?
Jeff C. -- Don't harm innocent bystanders.
_______________________________________________ Discuss mailing list Discuss@lists.surbl.org http://lists.surbl.org/mailman/listinfo/discuss
-
Chris -
Chris -
Daniel Kleinsinger -
Jeff Chan -
Rob McEwen