antispambox

Status

under development

container should be working basically

About

Antispambox is based on the idea of IMAPScan. It's an Docker container including ISBG. With ISBG it's possible to scan remotely an IMAP mailbox for SPAM mails with spamassassin. So we are not dependent to the SPAM filter of our provider.

Antispambox does have two anti-spam-engines integrated. Spamassassin and RSpamd.

Why not IMAPScan?

(Thanks to dc55028 for adding the MIT license to the IMAPScan repository)

I prefer Python instead of Bash scripts
I made several modifications (see Features) and not all of the modifications would be compatible to the ideas of IMAPScan
I integrated push support
...

Why not ISBG only?

ISBG is only supporting spamassassin as backend. Spamassassin is a very effective, but not very efficient SPAM filter. At home I'm running the docker container on a very small embedded PC with an Atom CPU. On my smartphone I'm using K9Mail with push support. So it is very important that the scanning for SPAM is very fast. With spamassassin it takes too long to filter the mails, so SPAM mails are shown on my smartphone before they are filterd out. The solution: rspamd.

But rspamd is not supported by ISBG and will not be supported to keep ISBG maintainable. So I forked ISBG and created IRSD.

Antispambox integrated both.

Why not IRSD only?

rspamd scans the mails on my machine very fast and efficient but the detection rate is not as good for me as spamassassin.

Features

Integration of ISBG (spamassassin)
Integration of IRSD (rspamd)
Integrated PUSH / IMAP IDLE support
integrated geo database and filters for it
focused on encrypted emails (header analysis only)
custom spamassassin rules for Germany and header analysis (my mails are prefiltered by mailbox.org - this container is only focused to the SPAM the MBO filter does not catch) - so the rules may not match your requirements
In future: Webinterface to configure everything
In future: Configure the rspamd webinterface
In future: Support multiple IMAP accounts within one container

Using the container

building the container

docker build -f Dockerfile -t antispambox . --no-cache

starting the container

sudo docker volume create bayesdb
sudo docker volume create accounts
sudo docker run -d --name antispambox --restart always -v bayesdb:/var/spamassassin/bayesdb -v accounts:/root/accounts antispambox

workflow and configuration

To configure the container run:
- docker exec -i -t antispambox /bin/bash
- use nano to configure the /root/accounts/imap_accounts.json
startup.py will be directly started with the docker container. To enable the scanning for spam, you need to set in /accounts/imap_accounts.json the enabled flag to True. By deafult this flag will be set to False until the configuration is finished.
First configure you mail account in /root/accounts/imap_accounts.json
Train spamassassinn and spamd:
- To ensure that spamassassin and spamd bayes filtering is working you should train it at least with 200 SPAM and 200 HAM mails.
  
  To train the backends copy your SPAM and HAM messages in the IMAP folders you configured for SPAM_train and HAM_traing.
Set the enabled flag in /root/accounts/imap_accounts.json to True.
Restart the docker container
The docker container will not start with IMAP idle to your INBOX folder and check for new mails. If a SPAM mail is detected, Antispambox will move the SPAM to your JUNK folder.
Mails you move manually to SPAM_train will be learned as SPAM. Mails you move manually to HAM_train will learned as HAM. The backend services spamassassin and rspamd will learn improve their detection rate with each learned mail.

Hints

To see how many mails rspamd has already learned or detected as SPAM or HAM, just run: spamc stat
To see how many mails spamassassin has already learned run: sa-learn --dump magic
There is a logfile to see the IMAP idle and scanning process: /var/log/antispambox.log

TODOs

see features
PEP8 & static code analysis

License

MIT

see license text

rsmuc/antispambox