Anybody tried CRM114 with ZCS

[uxbod]

The CRM114 Discriminator - The Controllable Regex Mutilator

Have just set it up so will see how it goes ...

john.doe@messagefortress.com

[uxbod]

So far so good Have set the scores low just in case. One FP but since retraining it has HAM have had not one bad hit. Will post stats soon and a how-to.

[uxbod]

==================================================================================
Spam Score Percentiles        0%       50%       90%       95%       98%      100%
----------------------------------------------------------------------------------
Score Spam (16)            9.984    26.106    31.878    36.790    43.123    47.345
Score Ham (30)            -4.264    -2.481     3.656     5.057     6.135     6.462
==================================================================================

======================================================================================================
Spam Score Frequency      <= -10     <= -5      <= 0      <= 5     <= 10     <= 20     <= 30      > 30
------------------------------------------------------------------------------------------------------
Hits (46)                      0         0        24         7         3         8         2         2
Percent of Hits            0.00%     0.00%    52.17%    15.22%     6.52%    17.39%     4.35%     4.35%
======================================================================================================

========================================================================================
SpamAssassin Rule Hits: Spam
----------------------------------------------------------------------------------------
Rank     Hits    % Msgs   % Spam    % Ham      Score Rule
----     ----    ------   ------    -----      ----- ----
   1       14    27.45%   87.50%    5.71%          1 CRM114_SPAM
   2       14    27.45%   87.50%   20.00%      0.001 HTML_MESSAGE
   3       13    25.49%   81.25%    0.00%          8 L_AV_SS_Spam
   4       13    25.49%   81.25%   14.29%        3.5 BAYES_99
   5       12    23.53%   75.00%    0.00%      1.955 URIBL_BLACK
   6       12    23.53%   75.00%    2.86%       1.96 RCVD_IN_BL_SPAMCOP_NET
   7       11    21.57%   68.75%    0.00%      1.501 URIBL_JP_SURBL
   8       10    19.61%   62.50%    2.86%      1.499 URIBL_SBL
   9        9    17.65%   56.25%    0.00%        1.5 RAZOR2_CF_RANGE_E8_51_100
  10        9    17.65%   56.25%    0.00%        0.5 RAZOR2_CF_RANGE_51_100
  11        9    17.65%   56.25%    0.00%        0.5 RAZOR2_CHECK
  12        7    13.73%   43.75%    0.00%        1.5 URIBL_OB_SURBL
  13        7    13.73%   43.75%    0.00%        3.7 PYZOR_CHECK
  14        5     9.80%   31.25%    0.00%        1.5 URIBL_WS_SURBL
  15        5     9.80%   31.25%    0.00%        1.5 RAZOR2_CF_RANGE_E4_51_100
  16        4     7.84%   25.00%    0.00%      1.083 URIBL_RHS_DOB
  17        4     7.84%   25.00%   11.43%          0 AV:MSRBL-Images.4-0-wx3c.UNOFFICIAL
  18        4     7.84%   25.00%    0.00%      0.001 DIGEST_MULTIPLE
  19        3     5.88%   18.75%    0.00%      1.457 MIME_HTML_ONLY
  20        3     5.88%   18.75%    5.71%      0.001 BAYES_50

Looks like it is doing the job. Set scores very low at the moment while it trains up. Once confident of no FPs will increase the scores.