Scoring questions

Computer Bob bob at inter-control.com
Mon Jan 29 21:23:34 CET 2018


Changes made, amavis restarted.
I have seen the following on all mails, I just was too lazy to include 
it because I had to blank the server name...skuza..

X-Virus-Scanned: Debian amavisd-new at M1-2.myorganization.org


On 1/29/18 2:15 PM, Dino Edwards wrote:
>
> Please try
>
> $sa_tag_level_deflt = undef;
>
> In
>
> /etc/amavis/conf.d/50-user
>
> Do you see the X-Virus-Scanned header in the emails that amavisd 
> processes?
>
> *From:*amavis-users 
> [mailto:amavis-users-bounces+dino.edwards=mydirectmail.net at amavis.org] 
> *On Behalf Of *Computer Bob
> *Sent:* Monday, January 29, 2018 2:40 PM
> *To:* amavis-users at amavis.org
> *Subject:* Re: Scoring questions
>
> I also agree that at this point auto learn should be off and cleared 
> as I have done.
> But I still continue to get garbage mails through showing headers such as:
>
> X-Spam-Flag: NO
> X-Spam-Score: 0.61
> X-Spam-Level:
> X-Spam-Status: No, score=0.61 tagged_above=-9999 required=5
>          tests=[HTML_FONT_LOW_CONTRAST=0.001, HTML_IMAGE_RATIO_04=0.61,
>          HTML_MESSAGE=0.001, SPF_HELO_PASS=-0.001, SPF_PASS=-0.001,
>          T_REMOTE_IMAGE=0.01, T_RP_MATCHES_RCVD=-0.01]
>          autolearn=no autolearn_force=no
>
> And as I said, when I run them through SA at the command line they 
> seem to score correctly.
> The scores being given in the headers can't be correct as they all are 
> similarly low and wrong.
> An interesting note is that if I try and forward one of these 
> received, they get flagged and sent to spam.
> Without knowing the intricacies of the amavis procedural steps, or 
> were to start, it is not possible for me to troubleshoot.
>
> On 1/29/18 1:20 PM, Dino Edwards wrote:
>
>     I disagree it's bad advice considering it's autolearn that seems to be creating at least some of the problems he's experiencing.
>
>     However, I do agree, the AutoLearn Threshold should definitely be set IF you are going to be using autolearn but in my experience auto-learn creates more problems than it solves. I believe that only humans should be be used for training the bayes database. Auto-learning has the tendency to exaggerate issues over time.
>
>     Keep it simple for now and train your bayes database and after you've trained it and it's scoring well, then consider using autolearn.
>
>     -----Original Message-----
>
>     From: amavis-users [mailto:amavis-users-bounces+dino.edwards=mydirectmail.net at amavis.org] On Behalf Of Benny Pedersen
>
>     Sent: Monday, January 29, 2018 1:06 PM
>
>     To:amavis-users at amavis.org <mailto:amavis-users at amavis.org>
>
>     Subject: Re: Re: Scoring questions
>
>     Computer Bob skrev den 2018-01-29 18:57:
>
>         I assume you mean bayes_auto_learn in local.cf. I set it to 0 from 1
>
>         and restarted.
>
>     yes its just bad advise, but setting this is what disables autolearn
>
>     i suggest see autolearnthreashold instaed
>
>     https://spamassassin.apache.org/full/3.3.x/doc/Mail_SpamAssassin_Plugin_AutoLearnThreshold.html
>
>     bayes_auto_learn_threshold_nonspam -5
>
>     bayes_auto_learn_threshold_spam 7.5
>
>     let the spammers win now :)
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.amavis.org/pipermail/amavis-users/attachments/20180129/5b390ade/attachment.html>


More information about the amavis-users mailing list