Spam

Introduction
This article describes the steps necessary to install and configure Exim on FreeBSD 6.2 with support for the following:

• authenticated SMTP (asmtp) using SASLAuthd
  
• spam detection and quarantine using SpamAssassin
  
• malware detection and quarantine using ClamAV

Each of the required 'dependencies' or components will be installed and configured, Exim will be installed and configured and finally we will test to check each component is working as required.

With regards to spam and malware scanning, the system described will quarantine any files/messages that it finds classified as spam or malware. In this way the quarantined files can be checked over by the admin at a later date and various stats gathering can be done if required.


Installing and Configuring SASLAuthd
SASLAuthd is an authentication daemon that can handle authentication requests from 3rd party applications such as Exim - generally for any application that can't directly access a system password database because of permission restrictions. In this case running Exim MTA as 'root' is a potential security risk, so exim runs as the 'mailnull' user on FreeBSD. Unfortunately this means Exim can't easily read the system password database to authenticate users who want to send mail via the server, which is where SASLAuthd comes in. Any requests for authentication with Exim are passed on to the SASLAuthd daemon which will then verify whether the user credentials are valid - if so, the email is delivered, if not, it's rejected.

1. Install SASLAuthd from the FreeBSD ports tree:
   
   
   CODE:
   root@win /root# cd /usr/ports/security/cyrus-sasl2-saslauthd/
   root@win /usr/ports/security/cyrus-sasl2-saslauthd# make install
   ...
   root@win /usr/ports/security/cyrus-sasl2-saslauthd# rehash
   
   
   
2. Configure SASLAuthd to run at boot.
   
   Edit /etc/rc.conf to include the following:
   
   
   CODE:
   saslauthd_enable="YES"
   saslauthd_flags="-a getpwent"
   
   
   Note:
   SASLAuthd will run using the 'getpwent' authentication mechanism with the flag above. This method uses the passwd file directly instead of using other means like kerberos or PAM. If you require another method, check the manpage for saslauthd.
   
   
3. Start the SASLAuthd daemon running:
   
   
   CODE:
   root@win /usr/ports/security/cyrus-sasl2-saslauthd# cd /usr/local/etc/rc.d
   root@win /usr/local/etc/rc.d# ./saslauthd start
   
   
   
4. (Optional) Test the SASLAuthd daemon:
   
   Substitute 'user' and 'pass' for the username and password of a user
   account on your system:
   
   
   CODE:
   root@win /usr/local/etc/rc.d# testsaslauthd -u user -p pass
   0: OK "Success."
   
   

Installing and Configuring SpamAssassin
SpamAssassin (SA) is one solution to the problem of spam. SA can run as a daemon (spamd) in the background and accept requests from an MTA such as Exim to check whether an email message should be classified as spam.

Spamd looks at the message and checks for various factors that make the message more or less likely to be spam and assigns the message a score based on what it finds. Spamd will then reply to the MTA, telling it the spam score that it gave that message. The MTA can then decide - based on that score - whether to accept/reject the message - or in the case of this guide whether to instead quarantine the message.

1. Install SA from the FreeBSD ports.
   
   Note:
   There are various installation options you can choose when installing SA which you should see when you first run 'make install' in the SA port directory. To see the options after already configuring them you can run 'make config'.
   
   In turn, each of SA's dependencies may also have options you can configure at install time.
   
   To write this guide I'm only using the single option 'AS_ROOT' in the SA install configuation and for the other items generally just choose the
   defaults.
   
   
   CODE:
   root@win /root# cd /usr/ports/mail/p5-Mail-SpamAssassin/
   root@win /usr/ports/mail/p5-Mail-SpamAssassin# make install
   
   
   Once complete, you should see:
   
   
   CODE:
   *************************************************************************
   *           _  _____ _____ _____ _   _ _____ ___ ___  _   _             *
   *          / \|_   _|_   _| ____| \ | |_   _|_ _/ _ \| \ | |            *
   *         / _ \ | |   | | |  _| |  \| | | |  | | | | |  \| |            *
   *        / ___ \| |   | | | |___| |\  | | |  | | |_| | |\  |            *
   *       /_/   \_\_|   |_| |_____|_| \_| |_| |___\___/|_| \_|            *
   *                                                                       *
   *       See /usr/local/share/doc/p5-Mail-SpamAssassin/INSTALL,          *
   *       and /usr/local/share/doc/p5-Mail-SpamAssassin/UPGRADE,          *
   *       or http://spamassassin.org/dist/INSTALL and                     *
   *       http://spamassassin.org/dist/UPGRADE BEFORE enabling            *
   *       this version of SpamAssassin for important information          *
   *       regarding changes in this version.                              *
   *************************************************************************
   
   
   It's a good idea to read the files listed in the banner above. SA has a large number of options that can be configured; a good place to start configuring options on FreeBSD is in /usr/local/etc/mail/spamassassin/.
   
   
2. Configure SA to run at boot.
   
   Edit /etc/rc.conf to include the following:
   
   
   CODE:
   spamd_enable="YES"
   
   
   
3. Start SA spamd.
   
   We can now go on to actually start spamd running as a daemon and verify spamd started ok:
   
   
   CODE:
   root@win /root# cd /usr/local/etc/rc.d
   root@win /usr/local/etc/rc.d# rehash
   root@win /usr/local/etc/rc.d# ./sa
   sa-spamd*   saslauthd*
   root@win /usr/local/etc/rc.d# ./sa-spamd start
   Starting spamd.
   munk@win /usr/local/etc/rc.d# ./sa-spamd status
   spamd is running as pid 754.
   
   
   This tells us spamd is running ok in the background.
   

Installing and Configuring ClamAV
ClamAV is an anti-virus suite and includes a daemon clamd (runs in the background to check for requests to test for virii), another daemon freshclam (updates the virus definition database) and a couple of clients to run on the commandline if you need them for local virus scanning.

Exim will send requests to the clamd server in much the same was as spamd does - if clamd classifies a message as containing a virus, Exim will reject delivery of the message and instaed quarantine it.

1. Install ClamAV from the FreeBSD ports tree:
   
   
   CODE:
   root@win /root# cd /usr/ports/security/clamav
   root@win /usr/ports/security/clamav# make install
   
   
   
2. Configure ClamAV to start at boot time.
   
   Edit /etc/rc.conf to include:
   
   
   CODE:
   clamav_clamd_enable="YES"
   clamav_freshclam_enable="YES"
   
   
   
3. Configure clamd.
   
   Edit /usr/local/etc/clamd.conf to include the following:
   
   
   CODE:
   LogFile /var/log/clamav/clamd.log
   PidFile /var/run/clamav/clamd.pid
   DatabaseDirectory /var/db/clamav
   LocalSocket /var/run/clamav/clamd
   FixStaleSocket
   User clamav
   AllowSupplementaryGroups
   ScanMail
   ScanArchive
   
   
   
4. Start clamd and freshclam.
   
   
   CODE:
   root@win /root# cd /usr/local/etc/rc.d
   root@win /usr/local/etc/rc.d# ./clamav-clamd start
   Starting clamav_clamd.
   root@win /usr/local/etc/rc.d# ./clamav-freshclam start
   Starting clamav_freshclam.
   
   
   Note:
   You may see the following message on first running clamd:
   
   
   CODE:
   LibClamAV Warning: **************************************************
   LibClamAV Warning: ***  The virus database is older than 7 days.  ***
   LibClamAV Warning: ***        Please update it IMMEDIATELY!       ***
   LibClamAV Warning: **************************************************
   
   
   As long as you're running freshclam, you can safely ignore this message. Freshclam should update your definitions automatically. Be sure to configure freshclam to update the virus definitions regularly.
   

ClamAV should be configured now and ready to accept request to check for malware from the Exim MTA.

We can now move on finally to install and configure Exim.


Installing and Configuring Exim
Exim configuration can be very complicated. This guide will only deal with the configuration of Exim so it accepts mail on a domain 'mail.example.com', scans the mail for malware/spam - quarantining anything it finds as malware/spam and accepts authentication requests correctly.

Important:
Ensure your mail server's DNS is configured correctly and preferably has a reverse DNS record (rDNS) set up. Many mail servers will not deliver mail correctly to/from your mail server without rDNS.

1. Install Exim from the FreeBSD ports tree:
   
   
   CODE:
   [12:10:57] root@win /root# cd /usr/ports/mail/exim
   [12:12:30] root@win /usr/ports/mail/exim#  make -DWITH_CONTENT_SCAN -DWITH_SASLAUTHD install
   
   
   
2. Stop the Sendmail daemon if it's already running:
   
   
   CODE:
   root@win /root# cd /etc/rc.d
   root@win /etc/rc.d# ./sendmail stop
   
   
   
3. Configure Exim to run at boot time.
   
   Edit /etc/rc.conf to include:
   
   
   CODE:
   sendmail_enable="NO"
   sendmail_submit_enable="NO"
   sendmail_outbound_enable="NO"
   sendmail_msp_queue_enable="NO"
   exim_enable="YES"
   
   
   This has the effect of disabling sendmail at boot time - the default FreeBSD MTA - and running Exim instead.
   
   
4. Configure mailer.conf to use Exim as the default MTA.
   
   Edit /etc/mail/mailer.conf to read:
   
   
   CODE:
   sendmail          /usr/local/sbin/exim
   send-mail         /usr/local/sbin/exim
   mailq             /usr/local/sbin/exim -bp
   newaliases        /usr/bin/true
   
   
   This will allow any FreeBSD base system mail related commands to use Exim instead of Sendmail.
   

Configuring Exim
We now move on to configuring Exim.

1. Set the primary hostname.
   
   Edit /usr/local/etc/exim/configure.
   
   Find and edit the 'primary_hostname' line for your domain:
   
   
   CODE:
   primary_hostname = example.com
   
   
   This configures Exim to accept mail primarily for the 'example.com' domain - ie foobar@example.com.
   
   
2. Find and edit the following lines to read:
   
   
   CODE:
   av_scanner = clamd:/var/run/clamav/clamd
   spamd_address = 127.0.0.1 783
   
   
   
3. Configure the malware and spam Access Control Lists (ACLs).
   
   How malware/spam checking works in this system:
   We add a check in the acl_check_data ACL for spam and malware. Exim will request each email is checked for spam/malware by the relevant daemon - spamd for spam, clamd for malware. If the message is classified as spam/malware by the relevant daemons, Exim will add a header to the message 'X-Quarantine-Me-Spam' (similar for malware).
   
   Later on when it comes to actually delivering (termed 'routing' in Exim terminology), we add two routers to test for the existence of the headers that are added in the acl_check_data ACL if a message is found to be spam/malware. If the headers are found by the malware/spam routers, the message is not delivered but instead copied to a quarantine location on disk.
   
   This quarantine location can then be checked later by an admin to check if anything is amiss - ie regular non spam/malware mail that should really have been delivered.
   
   Once you're satisfied the configuration is working as it should - ie after a few months of operation - and not finding false positives, you can change the malware/spam acl checks to just deny instead of adding the quarantine headers. Having said that, I still opt to just quarantine malware/spam and remove it at a later date.
   
   On to configuring the data ACL:
   
   Modify the acl_check_data ACL to read/include:
   
   
   CODE:
   acl_check_data:
   
     # Deny if the message contains a virus. Before enabling this check, you
     # must install a virus scanner and set the av_scanner option above.
     #
     # defer_ok - pass this message if scanner is down etc:
     warn  message         = X-Quarantine-Me-Malware: $malware_name
           log_message     = malware: $malware_name
           demime          = *
           malware         = */defer_ok
   
     # Add headers to a message if it is judged to be spam. Before enabling this,
     # you must install SpamAssassin. You may also need to set the spamd_address
     # option above.
     #
     warn    message       = X-Quarantine-Me-Spam: SA score $spam_score\n\
                             X-SA-Report: $spam_report
             log_message   = Spam score $spam_score > 5
             spam          = spamd/defer_ok
             condition     = ${if >{$spam_score_int}{50}{1}{0}}
   
     # Accept the message.
   
     accept
   
   
   At the top of the routers section, modify to read/include:
   
   
   CODE:
   begin routers
   
   check_malware:
     driver                = redirect
     condition             = ${if def:h_X-Quarantine-Me-Malware: {1}{0}}
     headers_add           = X-Quarantined-Malware: $h_X-Quarantine-Me-Malware:
     headers_remove        = X-Quarantine-Me-Malware
     data                  = /var/quarantine/malware/malware.$tod_logfile
     file_transport        = address_file
   
   check_spam:
     driver                = redirect
     condition             = ${if def:h_X-Quarantine-Me-Spam: {1}{0}}
     headers_add           = X-Quarantined-Spam: $h_X-Quarantine-Me-Spam:
     headers_remove        = X-Quarantine-Me-Spam
     data                  = /var/quarantine/spam/spam.$tod_logfile
     file_transport        = address_file
   
   no_more
   
   
   Modify the authenticators section to read:
   
   
   CODE:
   begin authenticators
   plain:
   driver = plaintext
   public_name = PLAIN
   server_condition = ${if saslauthd{{$2}{$3}}{1}{0}}
   
   login:
   driver = plaintext
   public_name = LOGIN
   server_prompts = "Username:: : Password::"
   server_condition = ${if saslauthd{{$1}{$2}}{1}{0}}
   
   
   Save the /usr/local/etc/exim/configuration file.
   
   
4. Create the quarantine directories and change ownership to mailnull:mail:
   
   
   CODE:
   root@win /root# mkdir -p /var/quarantine/{malware,spam}
   root@win /root# chown mailnull:mail /var/quarantine/{malware,spam}
   
   
   
5. Restart Exim to suck in the new config options:
   
   
   CODE:
   root@win /root# /usr/local/etc/rc.d/exim restart
   Stopping exim.
   Starting exim.
   
   

Exim should now be set to check for malware/spam and to authenticate users.


Testing Exim configuration
Finally we can move on to test that our config works correctly for spam/malware checking and for authenticating users.

1. Testing Exim's malware/spam scanning.
   
   The easiest option is to send an email to your mailserver with specially crafted malware/spam signatures included in the body of the message. When spamd/clamd see these signature strings in the body of the messages, they should classify the message as spam/malware and Exim in turn will quarantine the messages.
   
   The official EICAR malware/virus testing signature is as follows:
   
   
   CODE:
   X5O%@AP[4\PZX54(P^)7CC)7}$EICAR-STANDARD-ANTIVIRUS-TEST-FILEspamcH+H*
   
   See here for the official string:
   
   http://www.eicar.org/anti_virus_test_file.htm
   
   The official GTUBE spam testing signature is as follows:
   
   
   CODE:
   XJS*C4JDBQADN1.NSBN3*2IDNEN*GTUBE-STANDARD-ANTI-UBE-TEST-EMAIL*C.34X
   
   See here for the official string:
   
   http://spamassassin.apache.org/gtube/
   
   Note:
   Another option for testing spam/malware scanning is to run exim from the commandline using the command 'exim -bh 127.0.0.1'. This will run an SMTP session from the commandline (think telnet) and allow you to inject your own specially crafted message using the signatures above. This requires you enter a valid SMTP session, something like:
   
   
   CODE:
   HELO example.com
   MAIL FROM:foo@example.com
   RCPT TO:foo@example.com
   DATA
   X5O%@AP[4\PZX54(P^)7CC)7}$EICAR-STANDARD-ANTIVIRUS-TEST-FILEspamcH+H*
   .
   
   
   This would simulate the injection of a mail message with a virus in it and in 'exim -bh' mode you can see a lot of useful debugging info to verify everything works ok.
   
   
2. Testing Exim's Authentication configuration.
   
   We can now test that ASMTP is working. For this you can either run exim in one of it's many excellent debugging modes or you can simply configure a remote email client to use ASMTP. This guide will use the commandline to test ASMTP.
   
   Important:
   Before attempting this method please read the exim documentation on how ASMTP works. The following assumes you have read and understood that text.
   
   First create a simple perl script called 'encode' in /usr/local/etc/exim/ and make sure it is executable:
   
   
   CODE:
   root@win /usr/local/etc/exim# cat encode
   #!/usr/bin/perl
   use MIME::Base64;
   printf ("%s", encode_base64(eval ""$ARGV[0]""));
   root@darkstar /usr/local/etc/exim# chmod +x encode
   root@darkstar /usr/local/etc/exim# ls -al encode
   -rwxr-xr-x  1 root  wheel  85 Apr 23 12:25 encode
   
   
   Now decide which user account on your server you wish to test ASMTP with. It must be an account you know the password for obviously. I created an account called 'dummy' and set the password to 'dummy' as well - if you do this remember to remove the account or disable it as soon as you've finished testing.
   
   Encode the user:password pair into base64 MIME using the 'encode' script we created above:
   
   
   CODE:
   root@darkstar /usr/local/etc/exim# ./encode "\0dummy\0dummy"
   AGR1bW15AGR1bW15
   
   
   Now enter into Exim's fake SMTP session command-line mode and just for good measure do it in authentication debug mode as well:
   
   
   CODE:
   root@win /root#  exim -d+auth -bh 127.0.0.1
   Exim version 4.66 (FreeBSD 6.1) uid=0 gid=0 pid=3056 D=fbb95cfd
   Probably Berkeley DB version 1.8x (native mode)
   Support for: crypteq iconv() IPv6 use_setclassresources PAM Perl OpenSSL Content_Scanning Old_Demime
   Lookups: lsearch wildlsearch nwildlsearch iplsearch cdb dbm dbmnz dnsdb dsearch nis nis0 passwd
   Authenticators: cram_md5 plaintext spa
   Routers: accept dnslookup ipliteral manualroute queryprogram redirect
   Transports: appendfile/maildir/mailstore/mbx autoreply lmtp pipe smtp
   Fixed never_users: 0
   Size of off_t: 8
   changed uid/gid: forcing real = effective
     uid=0 gid=0 pid=3056
     auxiliary group list: 0
   seeking password data for user "mailnull": using cached result
   getpwnam() succeeded uid=26 gid=26
   seeking password data for user "root": cache not available
   getpwnam() succeeded uid=0 gid=0
   configuration file is /usr/local/etc/exim/configure
   log selectors = 00000ffc 00089001
   trusted user
   admin user
   changed uid/gid: privilege not needed
     uid=26 gid=6 pid=3056
     auxiliary group list: 6 6
   seeking password data for user "mailnull": cache not available
   getpwnam() succeeded uid=26 gid=26
   originator: uid=0 gid=0 login=root name=Charlie Root
   sender address = root@win.munk.me.uk
   sender_fullhost = [127.0.0.1]
   sender_rcvhost = [127.0.0.1]
   
   **** SMTP testing session as if from host 127.0.0.1
   **** but without any ident (RFC 1413) callback.
   **** This is not for real!
   
   host in hosts_connection_nolog? no (option unset)
   LOG: smtp_connection MAIN
     SMTP connection from [127.0.0.1]
   host in host_lookup? yes (matched "*")
   looking up host name for 127.0.0.1
   DNS lookup of 1.0.0.127.in-addr.arpa (PTR) succeeded
   IP address lookup yielded localhost.munk.me.uk
   gethostbyname2 looked up these IP addresses:
     name=localhost.munk.me.uk address=::1
     name=localhost.munk.me.uk address=127.0.0.1
   checking addresses for localhost.munk.me.uk
     ::1
     127.0.0.1 OK
   sender_fullhost = localhost.munk.me.uk [127.0.0.1]
   sender_rcvhost = localhost.munk.me.uk ([127.0.0.1])
   set_process_info:  3056 handling incoming connection from localhost.munk.me.uk [127.0.0.1]
   host in host_reject_connection? no (option unset)
   host in sender_unqualified_hosts? no (option unset)
   host in recipient_unqualified_hosts? no (option unset)
   host in helo_verify_hosts? no (option unset)
   host in helo_try_verify_hosts? no (option unset)
   host in helo_accept_junk_hosts? no (option unset)
   SMTP>> 220 win.munk.me.uk ESMTP Exim 4.66 Wed, 17 Jan 2007 19:24:22 +0000
   220 win.munk.me.uk ESMTP Exim 4.66 Wed, 17 Jan 2007 19:24:22 +0000
   smtp_setup_msg entered
   
   
   When you get to this point you are ready to start an SMTP 'conversation' with Exim. First introduce yourself to Exim using the SMTP 'EHLO localhost' command:
   
   
   CODE:
   EHLO localhost
   SMTP<< EHLO localhost
   sender_fullhost = localhost.munk.me.uk (localhost) [127.0.0.1]
   sender_rcvhost = localhost.munk.me.uk ([127.0.0.1] helo=localhost)
   set_process_info:  3103 handling incoming connection from localhost.munk.me.uk (localhost) [127.0.0.1]
   host in pipelining_advertise_hosts? yes (matched "*")
   host in auth_advertise_hosts? yes (matched "*")
   host in tls_advertise_hosts? no (option unset)
   250-win.munk.me.uk Hello localhost.munk.me.uk [127.0.0.1]
   250-SIZE 52428800
   250-PIPELINING
   250-AUTH PLAIN LOGIN
   250 HELP
   SMTP>> 250-win.munk.me.uk Hello localhost.munk.me.uk [127.0.0.1]
   250-SIZE 52428800
   250-PIPELINING
   250-AUTH PLAIN LOGIN
   250 HELP
   
   
   In response to your 'EHLO localhost' command, Exim returns more debug information but most importantly for us it also indicates what authentication options it offers in this line:
   
   
   CODE:
   250-AUTH PLAIN LOGIN
   
   
   This indicates that currently acceptable AUTH methods are PLAIN and LOGIN.
   
   We can then test the PLAIN login method using the "\0dummy\0dummy" user:password pair we encoded above:
   
   
   CODE:
   AUTH PLAIN AGR1bW15AGR1bW15
   SMTP<< AUTH PLAIN AGR1bW15AGR1bW15
   Running pwcheck authentication for user "dummy"
   pwcheck: success (NULL)
   plain authenticator:
     $1 =
     $2 = dummy
     $3 = dummy
   expanded string: 1
   SMTP>> 235 Authentication succeeded
   235 Authentication succeeded
   
   
   This indicates that authentication for 'dummy:dummy' would succeed and mail would be relayed (pending further conditional checks by Exim).
   

So we now have a working Exim with support for spam/malware checking and authentication over SMTP.

I just downloaded this great looking spam statistics plugin for Serendipity from Andreas. Unfortunately after installing it it didn't seem to work, so I got stuck in to see what was up.

Turns out it only works when the spamblock plugin logs to the database, so I'll either look into making it work with log files or maybe think about adding something to the admin stats plugin if that's possible. Or do neither given it's not uber important to me given I get a raft of info on the spam stats each night via a cron job.

I have a cron job that checks various things spam related on a daily basis - checking for referer spam, quarantined files uploaded via PHP, mod_security log entries that need attention and finally checking for serendipity / weblog spam. The situation with weblog spam had gotten so bad on the old domain munk.nu that I even ended up creating a script to convert spamblock log entries into firewall rules for ipf. I'm not kidding, at least 100 trackback spam entries per day through June and July - for the year 2006 so far there are nearly 9000 unique IPs dropping new trackback spam.

What's annoying too is that even adding offending IPs to my firewall block list, each and every new day there would be another 100 new unique IP addresses spamming the blog. No doubt this is a botnet - 100 new zombies found per day sounds like a professional organisation.

Ho hum. Anyway I'll add the 'log2ipf.pl' perl script in the extended part of this article. It's a perl script that's little more than an extended 'grep | sed' which searches for text in a file and then reports how many results it found for each item. In the default case using just 'log2ipf.pl somefile.log' it searches for:



in this case it reports a list of IP addresses and how many times each IP address was 'caught' trying to spam - but it could be modified to do anything. For example I have another 'filter' setup to see how many people use a google search to find pics on my server by searching for the term 'picasa.ini':



so I can feed apache logfiles to log2ipf.pl using this commandline:



telling me there was just 2 such searches during July 2006 (woo). I seem to remember that search returning more than that at the time I wrote the filter though lol. You get the idea anyway.

To add a new 'filter', best thing to do is import a sample logfile line you want to produce a result, then customize the script %re variable to include your custom filter.

For example, say you wanted to search for auth log failures for SSH (this is actually done for you by the periodic utility on FreeBSD if you set it up in /etc/periodic.conf, but that's another article! - you could write something like this for the %re filter:



which would result in:



As I say you can do the equivalent with grep, sed, sort and uniq on the commandline:



But for a very large file the timing differences between this method and the perl script are massive.

Anyhoo this is turning into a crazy long entry so I'll turn it in. The script log2ipf.pl - should rename that really since it's got little to do with ipf really! - is in the extended article below if anyone's interested.