rspamd and roundcube's markasjunk2 plugin

[David Mehler]

Hello,

Any users of rspamd and roundcube with the markasjunk2 plugin? If so,
do you have a configuration where the markasjunk2 plugin and rspamd
are talking to each other so users just have to select spam or not
spam on a message to train rspamd?

Thanks.
Dave.

[Владислав Ясюкевич]

Hello. It's easy to do.

Server side roundcube:

In /var/www/roundcube/plugins/markasjunk2/config.inc.php, add:

$config ['markasjunk2_spam_cmd'] = 'sudo /usr/local/bin/learn_spam %f %u';
$config ['markasjunk2_ham_cmd'] = 'sudo /usr/local/bin/learn_ham %f %u';

Where:

%u  -->  info@example.com
%l  -->  info
%d  -->  example.com
%i  -->  info@example.com
%f  -->  /var/www/roundcubemail-1.0.5markasjunk/temp/rcmSALearny62d3F

Script /usr/local/bin/learn_spam

#!/bin/bash -eu

declare -r FILE="$1"
declare -r MAIL="$2"

declare -r SERVER='*****'
declare -r DIR='/home/learn/mail'
declare -r MSGHASH="$(cat $FILE | md5sum | awk '{ print $1 }')"


ssh learn@${SERVER} "rm -f ${DIR}/HAM/${MAIL}.${MSGHASH}"
rsync -a --chmod=u+rw,g+rw,o+rw $FILE learn@${SERVER}:${DIR}/SPAM/${MAIL}.${MSGHASH}

Script /usr/local/bin/learn_ham

#!/bin/bash -eu

declare -r FILE="$1"
declare -r MAIL="$2"

declare -r SERVER='*****'
declare -r DIR='/home/learn/mail'
declare -r MSGHASH="$(cat $FILE | md5sum | awk '{ print $1 }')"


ssh learn@${SERVER} "rm -f ${DIR}/SPAM/${MAIL}.${MSGHASH}"
rsync -a --chmod=u+rw,g+rw,o+rw $FILE learn@${SERVER}:${DIR}/HAM/${MAIL}.${MSGHASH}

Server side rspamd:

In /etc/cron.d/learn

*/5 * * * *     root    /usr/local/bin/learn > /dev/null

Script /usr/local/bin/learn

#!/bin/bash -eu

declare -r HAMDIR='/home/learn/mail/HAM'
declare -r SPAMDIR='/home/learn/mail/SPAM'


[[ -d "$HAMDIR" ]] || {
    mkdir -p "$HAMDIR"
    chown -R learn: "$HAMDIR"
}

[[ -d "$SPAMDIR" ]] || {
    mkdir -p "$SPAMDIR"
    chown -R learn: "$SPAMDIR"
}


for HAMFILE in $(find "$HAMDIR" -type f); do
    HAMFILE_X="${HAMFILE%.*}"
    MAILNAME="${HAMFILE_X##*/}"
    [[ -n "${MAILNAME}" ]] && {
        rspamc -t 60 learn_ham --classifier=bayes_local-r "$MAILNAME" "$HAMFILE"

    }
    rm -f "$HAMFILE"
done


for SPAMFILE in $(find "$SPAMDIR" -type f); do
    SPAMFILE_X="${SPAMFILE%.*}"
    MAILNAME="${SPAMFILE_X##*/}"
    [[ -n "${MAILNAME}" ]] && {
        rspamc -t 60 learn_spam --classifier=bayes_local -r "$MAILNAME" "$SPAMFILE"
    }
    rm -f "$SPAMFILE"
done

For this template, you can make a different learning logic.

вторник, 10 апреля 2018 г., 5:52:30 UTC+5 пользователь David Mehler написал:

[Dedé Alves]

I'm using it.

Check this page:

https://wiki.dovecot.org/HowTo/AntispamWithSieve

There are some scripts to use with dovecot.

If you use IMAP, these scripts will be auto executed.

Regads,