You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@spamassassin.apache.org by Philipp Ewald <ph...@digionline.de> on 2019/10/22 10:21:45 UTC
Question about Bayes implementation
Hi folks,
at this point i split all my SPAM mail to get the attachment to create a
hash table. (but this is not my point)
Its also possible to split my SPAM into html/text, plain/text and
headers to.
Debian package: ripmime
Now i ask myself:
If i learn spamassassin with my mails should i learn with whole mail or
can i split them and learn only plain/text part? ore wich part would be
"the best" to learn?
thanks for help
kind regards
--
Philipp Ewald
Administrator
Re: Autolearn HAM with spamscore 996
Posted by John Hardin <jh...@impsec.org>.
On Tue, 22 Oct 2019, RW wrote:
> If you are in a position to train manually, I think it's best to
> turn-off auto-learning.
+1
Auto-learn is primarily for large sites with a diverse user base (e.g. an
ISP).
--
John Hardin KA7OHZ http://www.impsec.org/~jhardin/
jhardin@impsec.org FALaholic #11174 pgpk -a jhardin@impsec.org
key: 0xB8732E79 -- 2D8C 34F4 6411 F507 136C AF76 D822 E6E6 B873 2E79
-----------------------------------------------------------------------
The third basic rule of firearms safety:
Keep your booger hook off the bang switch!
-----------------------------------------------------------------------
936 days since the first commercial re-flight of an orbital booster (SpaceX)
Re: Autolearn HAM with spamscore 996
Posted by RW <rw...@googlemail.com>.
On Tue, 22 Oct 2019 18:31:18 +0200
Philipp Ewald wrote:
> First thanks for help, i will train them with current mail.
>
> my Amavis configuration found my Attachment and score this with SPAM
> score 999 but auto learn ignore this....
>
...
> did i miss something? can someone help me?
>
> google "auto learn amavis spamassassin" its really tricky to find
> something helpful.
There are some sanity checks on auto-learning, in this case there is the
rule that there has to be at least 3 points from header-based rules and
3 from body-based rules.
If you are in a position to train manually, I think it's best to
turn-off auto-learning.
Autolearn HAM with spamscore 996
Posted by Philipp Ewald <ph...@digionline.de>.
First thanks for help, i will train them with current mail.
my Amavis configuration found my Attachment and score this with SPAM
score 999 but auto learn ignore this....
X-Spam-Flag: YES
X-Spam-Score: 996.7
X-Spam-Level:
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
X-Spam-Status: Yes, score=996.7 tagged_above=-9999 required=5
tests=[AV:NSFW.UNOFFICIAL=999, RCVD_IN_DNSWL_MED=-2.3]
autolearn=ham autolearn_force=no
Test with GTUBE:
X-Spam-Flag: YES
X-Spam-Score: 997.7
X-Spam-Level:
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
X-Spam-Status: Yes, score=997.7 tagged_above=-9999 required=5
tests=[GTUBE=1000, RCVD_IN_DNSWL_MED=-2.3]
autolearn=no autolearn_force=no
Amavis config:
/etc/amavis/conf.d/50-user
@virus_name_to_spam_score_maps =
(new_RE( # the order matters!
[ qr'NSFW.UNOFFICIAL' => 999],
));
did i miss something? can someone help me?
google "auto learn amavis spamassassin" its really tricky to find
something helpful.
kind regards
Philipp
On 22.10.19 15:56, RW wrote:
>
> Train on the actual email.
>
--
Philipp Ewald
Administrator
Re: Question about Bayes implementation
Posted by RW <rw...@googlemail.com>.
On Tue, 22 Oct 2019 12:21:45 +0200
Philipp Ewald wrote:
> Hi folks,
>
> at this point i split all my SPAM mail to get the attachment to
> create a hash table. (but this is not my point)
>
> Its also possible to split my SPAM into html/text, plain/text and
> headers to.
> Debian package: ripmime
>
> Now i ask myself:
> If i learn spamassassin with my mails should i learn with whole mail
> or can i split them and learn only plain/text part? ore wich part
> would be "the best" to learn?
Train on the actual email.