Questions[spamassassin](server)

This is sort of a general question about training spamassassin. I have a newly set up mailserver which filters incoming mail through spamassassin. I recently got a flight reservation flagged as spam (score 5) and would like to tell spamassassin it's not spam. (Perhaps doing this would also re-send the mail without the modified spamassassin headers?)

I've tried searching around and am only finding stuff about either getting spamassassin to flag messages as spam (and not about fixing false positives), or for people writing emails - how not to be flagged as spam.

So in regards to giving spamassassin feedback on wrong calls:

1. Is there a way to do this from within an email client (for example: Thunderbird)

2. Is there a way to do this via the command-line on the mail server?

I'd like to make the process as fluid as possible, but whatever gets the job done.

Details from SpamAssassin regarding the email:

 0.0 FSL_HELO_NON_FQDN_1    No description available.
 0.6 HK_RANDOM_ENVFROM      Envelope sender username looks random
-0.0 RCVD_IN_DNSWL_NONE     RBL: Sender listed at http://www.dnswl.org/, no trust [82.150.225.129 listed in list.dnswl.org]
-0.0 RCVD_IN_MSPIKE_H3      RBL: Good reputation (+3) [82.150.225.129 listed in wl.mailspike.net]
 0.0 HEADER_FROM_DIFFERENT_DOMAINS From and EnvelopeFrom 2nd level mail domains are different
 1.0 SPF_SOFTFAIL           SPF: sender does not match SPF record (softfail)
 1.6 SUBJ_ALL_CAPS          Subject is all capitals
 1.1 MIME_HTML_ONLY         BODY: Message only has text/html MIME parts
 0.7 HTML_IMAGE_ONLY_20     BODY: HTML: images with 1600-2000 bytes of words
 0.0 HTML_MESSAGE           BODY: HTML included in message
-0.0 RCVD_IN_MSPIKE_WL      Mailspike good senders
 0.0 UNPARSEABLE_RELAY      Informational: message has unparseable relay lines
 0.0 T_REMOTE_IMAGE         Message contains an external image

Clearly the main culprits are the all-caps subject line SUBJ_ALL_CAPS and the MIME_HTML_ONLY (I guess, no text alternative).

The email was for a flight booking confirmation and the subject looked like this:

 Subject: JENNINGS/NICHOLAS KOSSOW MR 24 JAN MOF DPS

Headers:

X-Envelope-From: <[email protected]>
X-Envelope-To: <[email protected]>
Received: from mail1.amadeus.net (unknown)
    by 147-49-15-51.rev.cloud.scaleway.com(Postfix 3.1.0/8.13.0) with SMTP id unknown
    Fri, 20 Jan 2017 07:55:10 +0000
    (envelope-from <[email protected]>
Received: from obeap115 (nat-dns-mnp.amadeus.net [82.150.225.129])
    by mail1.amadeus.net (Postfix) with ESMTP id 3F7A9200042
    for <[email protected]>; Fri, 20 Jan 2017 07:55:10 +0000 (GMT)
From: [email protected]
TO: [email protected]
Message-ID: <CTS/GA/C50D54421A07/[email protected]>
FND-Request-ID: <CTS/GA/C50D54421A07/[email protected]>
Job-ID: 1
Subject: JENNINGS/NICHOLAS KOSSOW MR 24 JAN MOF DPS
Date: Fri, 20 Jan 2017 07:55:09 +0000
Content-Type: multipart/mixed; 
    boundary="----=_Part_191904_1900935199.1484898909762"
MIME-Version: 1.0

I have a VPS which is originally destined to be a webserver but some minimal mail capabilities are needed to be deployed as well, including sending and receiving as standalone server.

The current setup is the following:

• Postfix reveices the mail, the users are in virtual tables, stored in MySQL
• on connection all servers are tested with policyd-weight service against some DNSBLs
• all mail is runs through SpamAssassin spamd with the help of spamc client
• the mail is then delivered with Dovecot 2' LDA (local delivery agent), virtual users as well

As you saw...

• there's no virus scanner running, and that's for a reason: clamav eats all the memory possible and also, virus mails are all filtered out with this setup (I've tested the same with ClamAV enabled for 1,5 years, no virus mail ever got even to ClamAV)
• I don't use amavisd and I really don't want to. You only need that monster if you have plenty of memory and lots of simultaneous scanners. It's also a nightmare to fine tune by hand.
• I run policyd-weight instead of policyd and native DNSBLs in postfix. I don't like to send someone away because a single service listed them.

Important statement: everything works fine. I receive very small amount of spam, nearly never get a false positive and most of the bad mail is stopped by policyd-weight. The only "problem" that I feel the services at total uses a bit much memory alltogether.

I've already cut the modules of spamassassin (see below), but I'd really like to hear some advices how to cut the memory footprint as low as possible, mostly: what plugins SpamAssassin really needs and what are more or less useless, regarding to my current postfix & policyd-weight setup?

SpamAssassin rules are also compiled with sa-compile (sa-update runs once a week from cron, compile runs right after that)

These are some of the current configurations that may matter, please tell me if you need anything more.

postfix/master.cf (parts only)

dovecot   unix  -       n       n        -      -       pipe
  flags=DRhu user=vmail:vmail argv=/usr/bin/spamc -e /usr/lib/dovecot/deliver -d ${recipient} -f {sender}

postfix/main.cf (parts only)

smtpd_helo_required = yes
smtpd_helo_restrictions = permit_mynetworks,
        reject_invalid_hostname,
        permit

smtpd_recipient_restrictions =  permit_mynetworks,
        permit_sasl_authenticated,
        reject_invalid_hostname,
        reject_non_fqdn_hostname,
        reject_non_fqdn_recipient,
        reject_unknown_recipient_domain,
        reject_unauth_pipelining,
        reject_unauth_destination,
        check_policy_service inet:127.0.0.1:12525,
        permit

policyd-weight.conf (parts only)

$REJECTMSG = "550 Mail appeared to be SPAM or forged. Ask your Mail/DNS-Administrator to correct HELO and DNS MX settings or to get removed from DNSBLs";

$REJECTLEVEL = 4;
$DEFER_STRING = 'IN_SPAMCOP= BOGUS_MX=';
$DEFER_ACTION = '450';
$DEFER_LEVEL  = 5;
$DNSERRMSG = '450 No DNS entries for your MTA, HELO and Domain. Contact YOUR administrator';

# 1: ON, 0: OFF (default)
# If ON request that ALL clients are only checked against RBLs
$dnsbl_checks_only = 0;

# 1: ON (default), 0: OFF
# When set to ON it logs only RBLs which affect scoring (positive or negative)
$LOG_BAD_RBL_ONLY  = 1;

## DNSBL settings
@dnsbl_score = (
        # host,                 hit,    miss,   log name
        'dnsbl.ahbl.org',       3,      -1,     'dnsbl.ahbl.org',
        'dnsbl.njabl.org',      3,      -1,     'dnsbl.njabl.org',
        'dnsbl.sorbs.net',      3,      -1,     'dnsbl.sorbs.net',
        'bl.spamcop.net',       3,      -1,     'bl.spamcop.net',
        'zen.spamhaus.org',     3,      -1,     'zen.spamhaus.org',
        'pbl.spamhaus.org',     3,      -1,     'pbl.spamhaus.org',
        'cbl.abuseat.org',      3,      -1,     'cbl.abuseat.org',
        'list.dsbl.org',        3,      -1,     'list.dsbl.org',
);

# If Client IP is listed in MORE DNSBLS than this var, it gets REJECTed immediately
$MAXDNSBLHITS  = 3;

# alternatively, if the score of DNSBLs is ABOVE this level, reject immediately
$MAXDNSBLSCORE = 9;

$MAXDNSBLMSG = '550 Az levelezoszerveruk IP cime tul sok spamlistan talahato, kerjuk ellenorizze! / Your MTA is listed in too many DNSBLs; please check.';

## RHSBL settings
@rhsbl_score = (
        'multi.surbl.org',              4,      0,      'multi.surbl.org',
        'rhsbl.ahbl.org',               4,      0,      'rhsbl.ahbl.org',
        'dsn.rfc-ignorant.org',         4,      0,      'dsn.rfc-ignorant.org',
#       'postmaster.rfc-ignorant.org',  0.1,    0,      'postmaster.rfc-ignorant.org',
#       'abuse.rfc-ignorant.org',       0.1,    0,      'abuse.rfc-ignorant.org'
);

# skip a RBL if this RBL had this many continuous errors
$BL_ERROR_SKIP = 2;

# skip a RBL for that many times
$BL_SKIP_RELEASE = 10;

## cache stuff
# must be a directory (add trailing slash)
$LOCKPATH = '/var/run/policyd-weight/';

# socket path for the cache daemon.
$SPATH = $LOCKPATH.'/polw.sock';

# how many seconds the cache may be idle before starting maintenance routines
#NOTE: standard maintenance jobs happen regardless of this setting.
$MAXIDLECACHE = 60;

# after this number of requests do following maintenance jobs: checking for config changes
$MAINTENANCE_LEVEL = 5;

# negative (i.e. SPAM) result cache settings ##################################

# set to 0 to disable caching for spam results. To this level the cache will be cleaned.
$CACHESIZE = 2000;

# at this number of entries cleanup takes place
$CACHEMAXSIZE = 4000;

$CACHEREJECTMSG  = '550 temporarily blocked because of previous errors';

# after NTTL retries the cache entry is deleted
$NTTL = 1;

# client MUST NOT retry within this seconds in order to decrease TTL counter
$NTIME = 30;

# positve (i.,e. HAM) result cache settings ###################################

# set to 0 to disable caching of HAM. To this number of entries the cache will be cleaned
$POSCACHESIZE = 1000;

# at this number of entries cleanup takes place
$POSCACHEMAXSIZE = 2000;

$POSCACHEMSG = 'using cached result';

#after PTTL requests the HAM entry must succeed one time the RBL checks again
$PTTL = 60;

# after $PTIME in HAM Cache the client must pass one time the RBL checks again.
#Values must be nonfractal. Accepted time-units: s, m, h, d
$PTIME = '3h';

# The client must pass this time the RBL checks in order to be listed as hard-HAM
# After this time the client will pass immediately for PTTL within PTIME
$TEMP_PTIME = '1d';


## DNS settings

# Retries for ONE DNS-Lookup
$DNS_RETRIES = 1;

# Retry-interval for ONE DNS-Lookup
$DNS_RETRY_IVAL  = 5;

# max error count for unresponded queries in a complete policy query
$MAXDNSERR = 3;

$MAXDNSERRMSG = 'passed - too many local DNS-errors';

# persistent udp connection for DNS queries.
#broken in Net::DNS version 0.51. Works with Net::DNS 0.53; DEFAULT: off
$PUDP= 0;

# Force the usage of Net::DNS for RBL lookups.
# Normally policyd-weight tries to use a faster RBL lookup routine instead of Net::DNS
$USE_NET_DNS  = 0;

# A list of space separated NS IPs
# This overrides resolv.conf settings
# Example: $NS = '1.2.3.4 1.2.3.5';
# DEFAULT: empty
$NS  = '';

# timeout for receiving from cache instance
$IPC_TIMEOUT  = 2;

# If set to 1 policyd-weight closes connections to smtpd clients in order to avoid too many
#established connections to one policyd-weight child
$TRY_BALANCE  = 0;

# scores for checks, WARNING: they may manipulate eachother
# or be factors for other scores.
#  HIT score, MISS Score
@client_ip_eq_helo_score = (1.5, -1.25 );
@helo_score  = (1.5, -2 );
@helo_score  = (0, -2 );
@helo_from_mx_eq_ip_score= (1.5, -3.1  );
@helo_numeric_score= (2.5,  0 );
@from_match_regex_verified_helo= (1,-2 );
@from_match_regex_unverified_helo = (1.6, -1.5  );
@from_match_regex_failed_helo  = (2.5,  0 );
@helo_seems_dialup = (1.5,  0 );
@failed_helo_seems_dialup= (2, 0 );
@helo_ip_in_client_subnet= (0,-1.2  );
@helo_ip_in_cl16_subnet  = (0,-0.41 );
#@client_seems_dialup_score  = (3.75, 0 );
@client_seems_dialup_score  = (0, 0 );
@from_multiparted  = (1.09, 0 );
@from_anon= (1.17, 0 );
@bogus_mx_score = (2.1,  0 );
@random_sender_score  = (0.25, 0 );
@rhsbl_penalty_score  = (3.1,  0 );
@enforce_dyndns_score = (3, 0 );

spamassassin/init.pre (I've put the .pre files together)

loadplugin Mail::SpamAssassin::Plugin::Hashcash
loadplugin Mail::SpamAssassin::Plugin::SPF
loadplugin Mail::SpamAssassin::Plugin::Pyzor
loadplugin Mail::SpamAssassin::Plugin::Razor2
loadplugin Mail::SpamAssassin::Plugin::AutoLearnThreshold
loadplugin Mail::SpamAssassin::Plugin::MIMEHeader
loadplugin Mail::SpamAssassin::Plugin::ReplaceTags
loadplugin Mail::SpamAssassin::Plugin::Check
loadplugin Mail::SpamAssassin::Plugin::HTTPSMismatch
loadplugin Mail::SpamAssassin::Plugin::URIDetail    
loadplugin Mail::SpamAssassin::Plugin::Bayes
loadplugin Mail::SpamAssassin::Plugin::BodyEval
loadplugin Mail::SpamAssassin::Plugin::DNSEval
loadplugin Mail::SpamAssassin::Plugin::HTMLEval
loadplugin Mail::SpamAssassin::Plugin::HeaderEval
loadplugin Mail::SpamAssassin::Plugin::MIMEEval
loadplugin Mail::SpamAssassin::Plugin::RelayEval
loadplugin Mail::SpamAssassin::Plugin::URIEval
loadplugin Mail::SpamAssassin::Plugin::WLBLEval
loadplugin Mail::SpamAssassin::Plugin::VBounce
loadplugin Mail::SpamAssassin::Plugin::Rule2XSBody

spamassassin/local.cf (parts)

use_bayes                       1
bayes_auto_learn                1
bayes_store_module              Mail::SpamAssassin::BayesStore::MySQL
bayes_sql_dsn                   DBI:mysql:db:127.0.0.1:3306
bayes_sql_username              user
bayes_sql_password              pass
bayes_ignore_header             X-Bogosity
bayes_ignore_header             X-Spam-Flag
bayes_ignore_header             X-Spam-Status

### User settings
user_scores_dsn                 DBI:mysql:db:127.0.0.1:3306
user_scores_sql_password        user
user_scores_sql_username        pass
user_scores_sql_custom_query    SELECT preference, value FROM _TABLE_ WHERE username = _USERNAME_ OR username = '$GLOBAL' OR username = CONCAT('%',_DOMAIN_) ORDER BY username ASC

# for better speed
score DNS_FROM_AHBL_RHSBL       0
score __RFC_IGNORANT_ENVFROM    0
score DNS_FROM_RFC_DSN          0
score DNS_FROM_RFC_BOGUSMX      0
score __DNS_FROM_RFC_POST       0
score __DNS_FROM_RFC_ABUSE      0
score __DNS_FROM_RFC_WHOIS      0

UPDATE 01

As adaptr advised I remove policyd-weight and configured postfix postscreen, this resulted approximately -15-20 MB from RAM usage and a lot faster work. I'm not sure it's working at full capacity but it seems promising.

After some discussions with colleagues and the relevant depts. in my office, I have come to the conclusion that for my Postfix server, I must implement a before-queue filtering. The reason is that with an after-queue filtering, the email would be lost, and the sender (in case it's a false positive) will never realize that.

That is unacceptable for my company; if the falsely accused sender gets a notification, we want him/her to try re-sending the email through other channels. If he/she doesn't get a notification, he/she won't attempt a re-send.

Now, after perusing the various docs on SpamAssassin, I happened upon more than one way to implement SpamAssassin before-queue filtering. E.g.,

• SA + amavisd-new (run as smtpd proxy)
• SA + amavisd-new + amavisd-milter
• SA + MIMEdefang (run as milter)
• SA + spampd (run as smtpd proxy)
• SA + milter-spamd

and I'm sure you may also find other ways to implement it.

Which one(s) of the above is (/are) recommended?

Do you know of its pros/cons? Or perhaps do you have something better in mind?

I have a client that is getting heavily spammed.. It's the 15th of the month and POP3 bandwidth is almost 100 GB. There are only 7 e-mail accounts on this domain. I installed SpamAssassin set it to 5 and setup 10-20 filters reject most of the junk. I don't see much of a change in POP3 bandwidth. Correct me if I'm wrong, the server still receives the message using up bandwidth in order to analyze determine a spam score.

I stumbled across faking MX records, for thoes unaware--basically you set a bogus server as the lowest and highest MX records with the working server's MX record in the middle.

For example:

fake.example.com    1
realmx.example.com  2
fake2.example.com   3

The theory is, since majority of the spam is generated from Windows-based zombies and quite a few will query for the the highest MX record to spam since usually they're usually backup servers that don't filter spam. The lowest fake MX-record is for the rest of the spammers.. and generally spammers don't retry after failures.

Has anyone tried this? Does it help? Does it delay or cause issues with mail delivery? Does anyone else have a better solution?

I'm using SpamAssassin on Debian (the default configuration with Pyzor, AWL and Bayes disabled, and sa-compile enabled), and each of the spamd child processes consume around 100 to 150MB of memory (around 50MB of real memory) on the 32 bit servers, and about double this (logically enough) on the 64 bit servers. There are generally two child processes, but at busy times there can be five (the maximum) running.

ISTM that 200 to 600MB is a lot of memory for this task. I'd like to continue using SA as part of my filtering structure, but it's becoming difficult to justify so much memory.

Are there any ways to reduce the amount of memory that each child process uses? (Or alternatively, make a single child process so fast that I can set the maximum children to something like 2?). I'm willing to consider any options, including ones that will or may result in reduced accuracy.

I've already read the "Out of Memory Problems" page on the SA wiki; nothing there is of any use. Messages larger than 5 MB are not scanned with SA.