Re: spam control

2009-03-01 Thread Girish Kulkarni
Thanks for the all the replies!  I plan to use the Bogofilter to start
with and see how its comparison with Gmail's performance improves with
time.

Thanks,
Girish.

On Sat, Feb 28, 2009 at 9:29 AM, Girish Kulkarni gir...@hri.res.in wrote:
 I've been managing my mail with Gmail for about three years now.  The
 primary reason for sticking to Gmail is their spam control.

 I now want to try moving back to the old mail spool on my Debian box
 and start managing mail myself.  I'm curious to know what spam control
 techniques people on this list currently use.

 How is your experience with the spam control tools already available
 (SpamAssassin, SpanBouncer, Bogofilter)?  Which one is the most
 favoured?  Is Gmail indeed the best option for controlling spam?

 Thanks,
 Girish.

-- 
Girish Kulkarni - Allahabad, India - http://girish.50webs.com


--
To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org



Re: spam control

2009-02-28 Thread Tzafrir Cohen
On Sat, Feb 28, 2009 at 09:35:40AM +0200, Teemu Likonen wrote:

 Bogofilter is only a Bayesian-like filter so user have to teach it what
 is spam and what is not spam (ham). 

You probably have an existing spam collection (your spam folder at
gmail). As for ham: your other folders.

-- 
Tzafrir Cohen | tzaf...@jabber.org | VIM is
http://tzafrir.org.il || a Mutt's
tzaf...@cohens.org.il ||  best
ICQ# 16849754 || friend


-- 
To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org 
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org



Re: spam control

2009-02-28 Thread mouss
Girish Kulkarni a écrit :
 Hi,
 
 I've been managing my mail with Gmail for about three years now.  The
 primary reason for sticking to Gmail is their spam control.
 
 I now want to try moving back to the old mail spool on my Debian box
 and start managing mail myself.  I'm curious to know what spam control
 techniques people on this list currently use.
 
 How is your experience with the spam control tools already available
 (SpamAssassin, SpanBouncer, Bogofilter)?  Which one is the most
 favoured?  Is Gmail indeed the best option for controlling spam?
 

if you get mail directly (not relayed via any provider), then there are
controls you can do at smtp time. for example, with postfix:

reject_invalid_helo_hostname
reject_non_fqdn_helo_hostname
reject_rbl_client zen.spamhaus.org

These will block a lot of junk before data is even received. if you
decide to use postfix, you can get help on the postfix-users list. you
can get similar functionality in other MTAs.

for the spam that goes through, spamassassin is effective out of the
box. you can tune it for better results (use sa-update with selected
channels, train Bayes, ... etc). for more infos, the spamassassin-users
list is a good place to ask.

pure bayesian filters like bogofilter can be very effective if you
train them correctly. Depending on your mail flow, this may or may not
be easy. you can get help on the bogofilter list.


-- 
To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org 
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org



Re: spam control

2009-02-28 Thread Tzafrir Cohen
On Sat, Feb 28, 2009 at 09:29:10AM +0530, Girish Kulkarni wrote:

 How is your experience with the spam control tools already available
 (SpamAssassin, SpanBouncer, Bogofilter)?  Which one is the most
 favoured?  Is Gmail indeed the best option for controlling spam?

At work I'm forced to use gmail. As I don't like its interface
(including the strange imap semantics) I just pull everything via pop3
to my own server. At home I just use postifx/dovecot-imap. I use
spamassassin as the spam filter.

With most lists I filter in my fetchmailrc before calling to
spamassassin, as those lists rarely have spam. Debian lists seem to have
more spam than others[1] and thus I do filter them.

Even with spamc (the small client that connects to a running
spamassassin daemon) I can still get heavy load from several
simultanious assassinations. Hence I have:

  # The lock file ensures that only 1 spamassassin invocation happens
  # at 1 time, to keep the load down.
  #
  :0fw
  *  256000
  | /usr/bin/spamc

To make it simpler to search for mistakes, I filter the worst offenders
to a separate folder:

  :0
  * ^X-Spam-Level: \*\*\*\*\*\*\*\*\*\*\*\*\*\*\*
  .spam-bad/

  :0
  * ^X-Spam-Status: Yes
  .spam-maybe/

I use the 'autolearn' feature of SA (which is enabled by default). This
saves me most retraining, but I still need to correct errors.
If a spam does get through I move it to the folder 'spam' . In my
crontab I have a daily job that learns new spam from there.

  $ crontab -l
  # m h  dom mon dow   command
  10  4  *   *   * /home/tzafrir/bin/spam_learn

  $ cat /home/tzafrir/bin/spam_learn
  #!/bin/sh

  DIR=$HOME/Maildir/.spam
  DIR_NEW=$DIR/new
  DIR_CUR=$DIR/cur

  sa-learn --spam $DIR_NEW $DIR_CUR


[1] This is due to the policy of allowing anyone to post. I must say
that I'm surprised at just how little spam gets through - Admins of the
liszt-server are doing a great job. And I fully appreciate this policy 
and occasionally post to Debian lists I'm not subscribed to)

-- 
Tzafrir Cohen | tzaf...@jabber.org | VIM is
http://tzafrir.org.il || a Mutt's
tzaf...@cohens.org.il ||  best
ICQ# 16849754 || friend


-- 
To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org 
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org



Re: spam control

2009-02-28 Thread Andrew McGlashan

Hi,

Tzafrir Cohen wrote:

You probably have an existing spam collection (your spam folder at
gmail). As for ham: your other folders.


Only if he let's it collect or he captures and keeps it over time.

Kind Regards
AndrewM

Andrew McGlashan
Broadband Solutions now including VoIP


--
To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org 
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org




Re: spam control

2009-02-28 Thread mouss
Andrew McGlashan a écrit :
 Hi,
 
 Tzafrir Cohen wrote:
 You probably have an existing spam collection (your spam folder at
 gmail). As for ham: your other folders.
 
 Only if he let's it collect or he captures and keeps it over time.

and he (or a trusted person) _verified_ it. training using unqualified
mail is counter-productive.


-- 
To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org 
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org



Re: spam control

2009-02-27 Thread green
On Sat, 2009-02-28, 059, Girish Kulkarni wrote:
 How is your experience with the spam control tools already available
 (SpamAssassin, SpanBouncer, Bogofilter)?  Which one is the most
 favoured?  Is Gmail indeed the best option for controlling spam?

I pull all mail from Gmail, including spam, using IMAP and getmail4.  Then I 
deliver it to Postfix running locally which looks at ~/.forward and sees
| spamc | procmail.  So spamc (spamassassin) adds spam headers and procmail 
(I plan to switch to maildrop eventually) does some filtering (no spam 
filtering there).  I have left all the important spamassassin settings (like 
required_score) at the defaults.

I am inclined to measure spam-filtering success by the number of 
false-positives (heh, the ultimate spam filter filters nothing...!).  If I 
remember correctly, Gmail has had at 3-8 false positive while SpamAssassin has 
had 1-2 false positives.  If I take out mailing list messages, it drops to 1-2 
and 1 respectively.  I suppose that more mail has gone through SpamAssassin now 
than did Gmail when I was using it instead.

I do get a few false-negatives daily but not more than maybe 6; I don't 
remember how many with Gmail but I know I did get some.

Of course it is relevent how much mail I get in a day: I am subscribed to 
several mailing lists with debian-user being the highest traffic.


signature.asc
Description: Digital signature


Re: spam control

2009-02-27 Thread Andrew McGlashan

Hi,

Girish Kulkarni wrote:

I now want to try moving back to the old mail spool on my Debian box
and start managing mail myself.  I'm curious to know what spam control
techniques people on this list currently use.

How is your experience with the spam control tools already available
(SpamAssassin, SpanBouncer, Bogofilter)?  Which one is the most
favoured?  Is Gmail indeed the best option for controlling spam?


If you are in control of your incoming mail and can handle the cons of 
greylisting, then that would have to be one of the best methods to reduce 
bad emails.  I use greylisting with spamassassin to pretty good effect.


Kind Regards
AndrewM

Andrew McGlashan
Broadband Solutions now including VoIP


--
To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org 
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org




Re: spam control

2009-02-27 Thread Teemu Likonen
On 2009-02-28 09:29 (+0530), Girish Kulkarni wrote:

 How is your experience with the spam control tools already available
 (SpamAssassin, SpanBouncer, Bogofilter)? Which one is the most
 favoured? Is Gmail indeed the best option for controlling spam?

I have never used Gmail but I've been very happy with Bogofilter. I
fetch my mail with fetchmail and deliver it to mailboxes with
procmail. I have a Procmail rule to do Bogofilter check.

Bogofilter is only a Bayesian-like filter so user have to teach it what
is spam and what is not spam (ham). At first the spam filter doesn't
work at all. If you use Bogofilter I suggest configuring your mail
client so that it's easy to classify messages as spam or ham, that is,
to conveniently teach the filter. Gradually the filter gets better and
better and is optimized for the kind of mail you receive.


-- 
To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org 
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org



Re: spam control

2009-02-27 Thread Bob Cox
On Sat, Feb 28, 2009 at 09:29:10 +0530, Girish Kulkarni (gir...@hri.res.in) 
wrote: 

 I've been managing my mail with Gmail for about three years now.  The
 primary reason for sticking to Gmail is their spam control.
 
 I now want to try moving back to the old mail spool on my Debian box
 and start managing mail myself.  I'm curious to know what spam control
 techniques people on this list currently use.
 
 How is your experience with the spam control tools already available
 (SpamAssassin, SpanBouncer, Bogofilter)?  Which one is the most
 favoured?  Is Gmail indeed the best option for controlling spam?
 
If you have control over your incoming email server then you can
possibly get away without even using spamassassin.  Using postgrey and
postfix with postfix set to reject for things like invalid HELO,
non-fqdn hostname, sender using a dynamic IP and/or in various
blacklists, I manage to *reject* most unwanted stuff rather than
accepting it and then, later, rejecting it with spamassassin.  I like to
think that rejecting it keeps me off the spammers' lists because it is
seen as a delivery failure.

-- 
Bob Cox.  Stoke Gifford, near Bristol, UK.
Please reply to the list only.  Do NOT send copies directly to me.
Debian on the NSLU2: http://bobcox.com/slug/


-- 
To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org 
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org