You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@spamassassin.apache.org by "Lawrence @ Rogers" <la...@nl.rogers.com> on 2010/10/21 17:57:07 UTC

prevent rule from being considered for Bayes auto-learning

Hi,

I recall reading somewhere that there is a way to prevent a rule from 
being considered for Bayes auto-learning. I am trying to create a rule 
that hits upon some obvious spam that I am seeing, yet I want to make 
sure (for now) that any scores it assigns are not used for anything 
Bayes-related. I cannot seem to find any documentation on how to do this 
(Google doesn't help). I think it is something to do with setting a 
tflag, but any guidance would be appreciated.

Regards,

Lawrence Williams
LCWSoft
www.lcwsoft.com

Re: prevent rule from being considered for Bayes auto-learning

Posted by Jason Bertoch <ja...@i6ix.com>.
On 2010/10/21 12:17 PM, Michael Scheidell wrote:
> we decided that we didn't too much care to auto learn as 'not spam',
> emails sent from marketing companies, (because the reverse is true for
> auto learn ham) thus:
>
> aa_scores.cf:tflags RCVD_IN_DNSWL_HI net nice noautolearn
> aa_scores.cf:tflags RCVD_IN_DNSWL_MED net nice noautolearn
> aa_scores.cf:tflags RCVD_IN_DNSWL_LOW  net nice noautolearn
> aa_scores.cf:tflags RCVD_IN_RP_SAFE net nice noautolearn
> aa_scores.cf:tflags RCVD_IN_RP_CERTIFIED     net nice noautolearn

I actually filed a bug on this...

https://issues.apache.org/SpamAssassin/show_bug.cgi?id=6344

-- 
/Jason


Re: prevent rule from being considered for Bayes auto-learning

Posted by Michael Scheidell <mi...@secnap.com>.
On 10/21/10 11:57 AM, Lawrence @ Rogers wrote:
> Hi,
>
> I recall reading somewhere that there is a way to prevent a rule from 
> being considered for Bayes auto-learning. I am trying to create a rule 
> that hits upon some obvious spam that I am seeing, yet I want to make 
> sure (for now) that any scores it assigns are not used for anything 
> Bayes-related. I cannot seem to find any documentation on how to do 
> this (Google doesn't help). I think it is something to do with setting 
> a tflag, but any guidance would be appreciated.
>
you can prevent your rule from being considered in the DECISION as to if 
it will auto learn the tokens in the email with tflag noautolearn.

I don't know of any flag that would prevent the rule itself, so example:

rule1, hits 15 points
rule2(bayes) hits 4 points. total is 19 points.  if rule2 has 
noautolearn flag, then the 'do we auto learn this' score is only 15 points.
if your threshold is > 15.1, then the whole email is not considered for 
auto learning.

if rule3 hits 2 points, your total score is 21 points, but 'decision' 
delta is at 17 points now, and whole email is autolearned as spam.

we decided that we didn't too much care to auto learn as 'not spam', 
emails sent from marketing companies, (because the reverse is true for 
auto learn ham) thus:

aa_scores.cf:tflags RCVD_IN_DNSWL_HI net nice noautolearn
aa_scores.cf:tflags RCVD_IN_DNSWL_MED net nice noautolearn
aa_scores.cf:tflags RCVD_IN_DNSWL_LOW  net nice noautolearn
aa_scores.cf:tflags RCVD_IN_RP_SAFE net nice noautolearn
aa_scores.cf:tflags RCVD_IN_RP_CERTIFIED     net nice noautolearn
aa_scores.cf:tflags RCVD_IN_IADB_UT_CPR_MAT net nice noautolearn
aa_scores.cf:tflags RCVD_IN_IADB_UT_CPR_30           net nice noautolearn
aa_scores.cf:tflags RCVD_IN_IADB_UT_CPEAR            net nice noautolearn
aa_scores.cf:tflags RCVD_IN_IADB_UNVERIFIED_2        net nice noautolearn
aa_scores.cf:tflags RCVD_IN_IADB_UNVERIFIED_1        net nice noautolearn
aa_scores.cf:tflags RCVD_IN_IADB_SPF                 net nice noautolearn
aa_scores.cf:tflags RCVD_IN_IADB_SENDERID            net nice noautolearn
aa_scores.cf:tflags RCVD_IN_IADB_RDNS                net nice noautolearn
aa_scores.cf:tflags RCVD_IN_IADB_OPTIN_LT50          net nice noautolearn
aa_scores.cf:tflags RCVD_IN_IADB_OPTIN_GT50          net nice noautolearn
aa_scores.cf:tflags RCVD_IN_IADB_OPTIN        net nice noautolearn
aa_scores.cf:tflags RCVD_IN_IADB_ML_DOPTIN        net nice noautolearn
aa_scores.cf:tflags RCVD_IN_IADB_MI_CPR_MAT        net nice noautolearn
aa_scores.cf:tflags RCVD_IN_IADB_MI_CPR_30        net nice noautolearn
aa_scores.cf:tflags RCVD_IN_IADB_MI_CPEAR        net nice noautolearn

> Regards,
>
> Lawrence Williams
> LCWSoft
> www.lcwsoft.com


-- 
Michael Scheidell, CTO
o: 561-999-5000
d: 561-948-2259
ISN: 1259*1300
 >*| *SECNAP Network Security Corporation

    * Certified SNORT Integrator
    * 2008-9 Hot Company Award Winner, World Executive Alliance
    * Five-Star Partner Program 2009, VARBusiness
    * Best in Email Security,2010: Network Products Guide
    * King of Spam Filters, SC Magazine 2008

______________________________________________________________________
This email has been scanned and certified safe by SpammerTrap(r). 
For Information please see http://www.secnap.com/products/spammertrap/
______________________________________________________________________  

Re: prevent rule from being considered for Bayes auto-learning

Posted by "Lawrence @ Rogers" <la...@nl.rogers.com>.
On 21/10/2010 2:17 PM, Karsten Bräckelmann wrote:
> On Thu, 2010-10-21 at 18:39 +0200, Karsten Bräckelmann wrote:
>> See M::SA::Plugin::AutoLearnThreshold. In a nutshell,  (a) there are a
>> few tflags that will prevent a rule's score to be used for auto-learning
>> and  (b) the score used is picked from the respective non-bayes
>> score-set.
>>
>> With (a) you can make a rule invisible to the auto-learning decision.
>> And by setting the scores for score-set 0 and 1 both to 0 as per (b),
>> you can effectively disable a rule unless Bayes is enabled.
> ... *and* have that rule "ignored" for the auto-learning decision, if
> Bayes and auto-learn is enabled. (Actually not ignored, but adding zero
> doesn't influence the result. ;)
>
> The tflags way is much more straight forward, though.
>
>
>> You cannot, however, create a rule to conditionally prevent auto-
>> learning altogether (which, as I understand isn't what you had in mind
>> anyway).
Thanks everyone, I have set the rule to noautolearn using the tflags 
directive (this is what I wanted, for the rule to simply not be 
considered when auto-learning).

- Lawrence

Re: prevent rule from being considered for Bayes auto-learning

Posted by Karsten Bräckelmann <gu...@rudersport.de>.
On Thu, 2010-10-21 at 18:39 +0200, Karsten Bräckelmann wrote:
> See M::SA::Plugin::AutoLearnThreshold. In a nutshell,  (a) there are a
> few tflags that will prevent a rule's score to be used for auto-learning
> and  (b) the score used is picked from the respective non-bayes
> score-set.
> 
> With (a) you can make a rule invisible to the auto-learning decision.
> And by setting the scores for score-set 0 and 1 both to 0 as per (b),
> you can effectively disable a rule unless Bayes is enabled.

... *and* have that rule "ignored" for the auto-learning decision, if
Bayes and auto-learn is enabled. (Actually not ignored, but adding zero
doesn't influence the result. ;)

The tflags way is much more straight forward, though.


> You cannot, however, create a rule to conditionally prevent auto-
> learning altogether (which, as I understand isn't what you had in mind
> anyway).

-- 
char *t="\10pse\0r\0dtu\0.@ghno\x4e\xc8\x79\xf4\xab\x51\x8a\x10\xf4\xf4\xc4";
main(){ char h,m=h=*t++,*x=t+2*h,c,i,l=*x,s=0; for (i=0;i<l;i++){ i%8? c<<=1:
(c=*++x); c&128 && (s+=h); if (!(h>>=1)||!t[s+h]){ putchar(t[s]);h=m;s=0; }}}


Re: prevent rule from being considered for Bayes auto-learning

Posted by Karsten Bräckelmann <gu...@rudersport.de>.
On Thu, 2010-10-21 at 13:27 -0230, Lawrence @ Rogers wrote:
> I recall reading somewhere that there is a way to prevent a rule from 
> being considered for Bayes auto-learning. I am trying to create a rule 
                       ^^^^^ ^^^^^^^^^^^^^
> that hits upon some obvious spam that I am seeing, yet I want to make 
> sure (for now) that any scores it assigns are not used for anything 
> Bayes-related. I cannot seem to find any documentation on how to do this 
> (Google doesn't help). I think it is something to do with setting a 
> tflag, but any guidance would be appreciated.
  ^^^^^

Yup, that's correct. Though your google-fu today... The three marked
strings from your own description leads to perfect documentation. :)

See M::SA::Plugin::AutoLearnThreshold. In a nutshell,  (a) there are a
few tflags that will prevent a rule's score to be used for auto-learning
and  (b) the score used is picked from the respective non-bayes
score-set.

With (a) you can make a rule invisible to the auto-learning decision.
And by setting the scores for score-set 0 and 1 both to 0 as per (b),
you can effectively disable a rule unless Bayes is enabled.

You cannot, however, create a rule to conditionally prevent auto-
learning altogether (which, as I understand isn't what you had in mind
anyway).


-- 
char *t="\10pse\0r\0dtu\0.@ghno\x4e\xc8\x79\xf4\xab\x51\x8a\x10\xf4\xf4\xc4";
main(){ char h,m=h=*t++,*x=t+2*h,c,i,l=*x,s=0; for (i=0;i<l;i++){ i%8? c<<=1:
(c=*++x); c&128 && (s+=h); if (!(h>>=1)||!t[s+h]){ putchar(t[s]);h=m;s=0; }}}


RE: prevent rule from being considered for Bayes auto-learning

Posted by Kevin Miller <Ke...@ci.juneau.ak.us>.
Lawrence @ Rogers wrote:
> Hi,
> 
> I recall reading somewhere that there is a way to prevent a rule from
> being considered for Bayes auto-learning. I am trying to create a
> rule that hits upon some obvious spam that I am seeing, yet I want to
> make sure (for now) that any scores it assigns are not used for
> anything Bayes-related. I cannot seem to find any documentation on
> how to do this (Google doesn't help). I think it is something to do
> with setting a tflag, but any guidance would be appreciated.      

I think you're looking for this:
  tflags          YOUR_RULENAME       noautolearn

HTH...

...Kevin
-- 
Kevin Miller                Registered Linux User No: 307357
CBJ MIS Dept.               Network Systems Admin., Mail Admin.
155 South Seward Street     ph: (907) 586-0242
Juneau, Alaska 99801        fax: (907 586-4500