Re: spam scores being neutralized by bayes

Derek Atkins <warlord-DPNOqEs/[email protected]>
Newsgroups gmane.mail.virus.maiamailguard
Message-ID <[email protected]>
Ed <[email protected]> writes:

> As I read it, the issue was SPAM was being scored as NOT SPAM because
> of the negative BAYES_00 score.  In this case you'd have to teach
> BAYES that it IS spam.
>
> My route takes care of the teaching issue as the bayes scores are
> transient because they are refreshed on a regular basis lest your
> database become ginormous before too long.
>
> My method can also fix scores of other non-bayes tests.
>
> And you're right BAYES_00 is rare, so fix it at some benign value like 0.

No, it's not rare at all if you train it correct.  It's just rare to
have a SPAM message meet BAYES_00.  I have HAM messages that reach
BAYES_00 all the time, so I'd LIKE a BAYES_00 score to give a lot of
weight.

> --Ed

-derek

> On 07/07/2015 11:25 AM, Derek Atkins wrote:
>> Or better yet, you could just teach the Bayes filter that it's not spam!
>> :)  Rarely do I find a spam message that matches the BAYES_00 rule.
>>
>> -derek
>>
>> Ed <[email protected]> writes:
>>
>>> Hi,
>>>
>>> In local.cf set your own score :
>>>
>>> score BAYES_00  0.0    ##  pick any number you want here
>>>
>>> Any scores you set in local.cf over ride any built in scoring for any token.
>>>
>>> --Ed
>>>
>>> On 07/07/2015 05:43 AM, Jason Street wrote:
>>>> I hope are small group of Maia users are keeping well.
>>>>
>>>> I am slowly getting to grips with Maia one problem I am having is with
>>>> the scores being assigned to email and as a result a lot of genuine spam
>>>> is reaching inboxes.
>>>>
>>>> The problem seems to be that Bayes (?) is neutralizing the scoring and
>>>> as a result email that would have been spam is allowed through.
>>>>
>>>> For example this scoring for spam that reached inbox
>>>>
>>>>     1.700     URIBL_DBL_SPAM     Contains an URL listed in the DBL blocklist
>>>>     1.449     RCVD_IN_BRBL_LASTEXT     (No description provided)
>>>>     0.437     HTML_IMAGE_RATIO_02     HTML has a low ratio of text to
>>>> image area
>>>>     0.100     DKIM_SIGNED     Domain Keys Identified Mail: message has a
>>>> signature
>>>>     0.001     HTML_MESSAGE     HTML included in message
>>>>     -0.001     SPF_PASS     SPF: sender matches SPF record
>>>>     -0.100     DKIM_VALID     Message has at least one valid DKIM or DK
>>>> signature
>>>>     -0.100     DKIM_VALID_AU     Message has a valid DKIM or DK signature
>>>> from author's domain
>>>>     -1.900     BAYES_00     Bayesian spam probability is 0 to 1%:)
>>>>
>>>> What is the best way to resolve this?
>>>>
>>>> ¬¬JS
>>>> _______________________________________________
>>>> Maia-users mailing list
>>>> [email protected]
>>>> http://www.renaissoft.com/cgi-bin/mailman/listinfo/maia-users
>>

-- 
       Derek Atkins, SB '93 MIT EE, SM '95 MIT Media Laboratory
       Member, MIT Student Information Processing Board  (SIPB)
       URL: http://web.mit.edu/warlord/    PP-ASEL-IA     N1NWH
       [email protected]                        PGP key available
_______________________________________________
Maia-users mailing list
[email protected]
http://www.renaissoft.com/cgi-bin/mailman/listinfo/maia-users
lmpx.com only provides a reader for public news (NNTP) servers. It is not affiliated with the servers or forums shown here and is not responsible for the content of articles, which is written by their respective authors.