http://sourceforge.net/projects/phpop3clean/


Download:
Support:
http://support.silisoftware.com

phPOP3clean.changelog.txt

/////////////////////////////////////////////////////////////////
/// phPOP3clean() by James Heinrich <info@silisoftware.com>    //
//  available at http://phpop3clean.sourceforge.net            //
/////////////////////////////////////////////////////////////////


--------------------
What is phPOP3clean?
--------------------

phPOP3clean is a PHP-based POP3 email scanner. It's designed to be
run as a cron job every minute or so, and to catch & delete several
types of unwanted emails:

a) malformed emails - incomplete or malformed headers, which cause
   some POP3 servers to drop connection when the message is retrieved

b) email worms - attached executable files matched against database
   of known variant, including matching variable-length files or
   files with internal random bytes (such as the currently-popular
   Netsky & Beagle variants).  Zipped attachments are unzipped and
   scanned. Password-protected zipped attachments are matched based
   on deceptive filenames (eg: "readme.txt   .exe").

c) image-based spam - attached images are matched against database of
   known spam images to reject messages containing only an inline
   attached image (technique of bypassing many spam filters). Images
   with random bytes appended are also matched.

d) obfuscated word spam - scans message body for obfuscated words,
   such as "víàqrä" in place of "viagra"

e) blacklisted phrase spam - scans message body for phrases (such as
   "Securities Exchange Act of 1934" or "forward looking statements",
   both of which are in most stock-promoting spam). Regular expression
   matches can be used to match variations.

f) blacklisted source code - scans message source for phrases known
   to be part of exploits (eg: <script language="JScript.Encode">)

g) blacklisted Received header - reject messages based on "Received"
   header contents

h) blacklisted IP spam - scans message contents for links to blacklisted
   IP ranges (eg: 221.11.133.66/25). Links can be in HTML or plain text,
   image/iframe src, etc.

i) blacklisted domains - auto-blacklists any IPs associated with domains
   that regularly swap IPs from a pool of zombie machines

j) whitelist - "From" and "Return-Path" headers are scanned to match
   whitelist to bypass all filtering.

k) SpamAssassin support - can delete emails based on what SpamAssassin says

m) DNSBL support - reject email based on headers or body containing
   blacklisted IPs/domains


All matching is done against MySQL tables, the contents of which are all
user-configurable with included admin interface.

Supported message encodings are: 7-bit, 8-bit, quoted-printable, base64.



------------
Requirements
------------

PHP v4.?.?
MySQL v3.x+


------------
Installation
------------

Unzip to a webroot (or subdirectory of your choice) on your server.
phPOP3clean.admin.php should be public-accessible and the /phPOP3clean/
subdirectory should be secured (.htaccess or similar).  The two
subdirectories ("md5imagecache" and "quarantine") should be made
writeable (chmod 777 will work, lower permission may work depending on
your server configuration).

For speed reasons it's advisable to run phPOP3clean on the same server
as the mailserver, but it works over POP3 so you can run phPOP3clean on
any webserver and scan accounts on any other server(s).


There are some values you must modify in phPOP3clean.config.php -- take a
look at that file and it should be pretty self-explanatory.

To create the MySQL tables required by phPOP3clean, simply run
phPOP3clean.install.php and the tables will be created if required. Any
changes to the table structures required by future versions will be
handled by this file, so run this again after upgrading to a newer
version of phPOP3clean.

A PHPOP3CLEAN_QUARANTINE directory (default is phPOP3clean/quarantine/
below installation directory) and within that a new directory is created
each month where the deleted emails are stored (gzipped). This allows you
to review deleted emails from the admin interface. You will need to
manually clean up these directories as the months go by.


After you've configured phPOP3clean, schedule it to run as a cron job
every few minutes (every minute is ideal, if your server can handle the
load.  Note: not every account is scanned every pass of phPOP3clean.php,
the scan frequency is configurable per account). The cron job may look
something like this:

  php -q ./httpdocs/phPOP3clean/phPOP3clean.php >> /dev/null

For improved security you can place the phPOP3clean files outside the web-
accessible DOCUMENT_ROOT.

Alternately, if your server does not support executing PHP files directly,
you can put phPOP3clean into a .htaccess-secured directory and call Lynx
to load the file:

  lynx -dump -auth=user:pass http://example.com/phPOP3clean/phPOP3clean.php

where user:pass is a valid username/password in the .htaccess file.


Note: phPOP3clean is an email-filtering framework only, by default it
will not delete much. What it will delete "out of the box" are emails:
* with malformed headers and no body (these sometimes crash servers)
* with very suspiciously-named attachments (eg: readme.txt.pif)
* containing IPs or domains listed in a DNS blacklist (if configured)

Beyond that, it's up to you to supply rules as to what should or should not
be considered deletable. However, I do release "definitions" every so often
(weekly-to-monthly) which contain banned words, phrases, code, images, etc
that I consider generic enough to work for everyone. The creation &
maintenance of an IP blacklist is, however, more personal so I leave that
entirely up to you (but the DNS-BlackList does automatically catch a large
number of undesirable emails).  To be effective, the manual IP blacklist
must be updated daily, if not hourly. Unfortunate, but true.


--------------
Administration
--------------

Adminstration should be simple and intuitive with the supplied admin
interface -- simply access phPOP3clean.admin.php and follow the directions.
Log into the admin page either with PHPOP3CLEAN_DBUSER + PHPOP3CLEAN_DBPASS
(full access) or a valid configured POP3 email address + password (for more
limited access).


--------------------------------
Administration - Update Database
--------------------------------

Every so often new virus/image definitions and word lists will be released
and you can easily update your installation to include these definitions.
The updates are released as zipped SQL files, consisting of REPLACE INTO
statements. To install the updates, simply click on "Update Database" in
the admin interface, browse to the unzipped SQL file and click "Upload &
Process".

There are 3 SQL updates available (from SourceForge):
  * phpop3clean_sql_nodata
  * phpop3clean_sql_fulldata-image
  * phpop3clean_sql_fulldata-exe
The "nodata" version is what everybody should download and update with. It
contains updates to the word lists, virus definitions and image definitions
but it does not contain the actual binary data for the virus and image
definitions. This means that you cannot see the image preview for attached
images, or download a sample of the virus/worms in the database. The MD5
hash will still match the unwanted email content and work fine, so there is
no compelling reason to install anything other than the "nodata" update.

The distribution of phPOP3clean does not come with any database content, so
you will need to download and install the latest definitions before using
phPOP3clean.


----------------------
Administration - Users
----------------------

You can add users to the system in two ways: the admin interface and public
signup. The admin interface has "User admin" which should be simple to use.
There is also phPOP3clean.demo.signup.php which provides a sample of how
to allow users to sign up for email filtering using a publicly-accessible
page on your site.


-------------------------------------
Administration - Infected Attachments
-------------------------------------

Any email attachments with executable file extensions (cmd, bat, vbs, cpl,
hta, pif, scr, com, exe) will be matched against the database of known
infected files. Many email worms (Netsky, Beagle, etc) append a few bytes
of random data to the end of the emailed file, and randomize a few bytes
in the middle of the file. phPOP3clean gets around this by truncating the
file before the appended random data, and setting the internal randomized
bytes to 0x00 before taking the MD5 hash of the file. The matching pattern
string has a format similar to this:

  17440|144-146;204;489

where "17440" is the length to which the file should be truncated (after
this offset is appended random data), and after the "|" comes a semicolon-
seperated list of bytes to be set to 0x00 before hashing. The list of bytes
can be either single bytes, or a hyphenated range of bytes.

Virus/worm names in phPOP3clean are formatted to match Symantec's naming
conventions, and links are provided to the SARC database.

Any attachments with executable file extensions that do not match anything
in the database will be saved in the database and an alert will be sent to
the administrative email address.

If you come across any infected attachments that are not caught by the
latest phPOP3clean definitions, please forward a copy of the infected file
to me at info@silisoftware.com and I will make sure to include it in the
next release.


--------------------------------
Administration - Attached Images
--------------------------------

The attached images processing is very similar to the Infected Attachments,
except only images are processed. An image preview is available in the
admin interface, as well as a description field. No images are
automatically added to the database, but you can upload images yourself.


-----------------------------
Administration - IP Blacklist
-----------------------------

phPOP3clean extract all URLs from emails and does a lookup on all the
domains contained in each email, and if any of the IPs match a blacklisted
IP then the email is deleted. To add IPs to the blacklist, click on "IPs
Blacklist" in the admin interface, and the copy-paste as many URLs (one per
line) into the textarea as you wish, and click "Add". The domains will be
extracted from the URLs, and split into subdomains and a lookup performed
on each. For example, "http://www.sub.example.com" would result in
three lookups: "example.com", "sub.example.com", "www.sub.example.com"
There is a DHTML display as the IP lookups are performed:


Once the IPs have been looked up, their status with respect to the current blacklist is displayed:

If the IP range is not blacklisted, a link will be provided to create a new blacklist (highlighted with red). If the IP is not blacklisted, but another IP that's very similar (ex: 127.0.0.1 vs 127.0.0.2) is in the blacklist, a link is provided to edit the blacklist. In the blacklist edit screen

you can change the dropdown mask depth to include all the desired IPs in that range. 32 gives no range, and 24 matches anything in the last quarter of the IP. The IP range for each value is displayed immediately below the dropdown. The textarea description is optional, and allows you to add comments, for example what domains are in this range. Whitelist feature: You can create a whitelist exactly the same as a blacklist. Whitelisting an IP range does not explicitly allow anything through; all it does is prevent you from accidentally blacklisting a known-good IP range. -------------------------------------------------- Administration - Words (clean, obfuscated, source) -------------------------------------------------- The contents of the email is scanned for three different types of phrases: * "clean" phrases, which appear exactly as entered in the message body * "obfuscated" phrases, which are human-readable but disguised (ex: viàgra) * "source code" phrases, mostly to scan for exploits and phishing Clean words/phrases are scanned for in the text-only portion of the message body, which may include the plaintext portion of the email, as well as the HTML version after the HTML tags have been stripped out. Obfuscated words/phrases are also scanned for in the message body. These are words/phrases that have characters substituted with similar ones, or letter repeated, so that the word is readable to a human but doesn't match exactly (a common technique to try and get around spam filters). Letters are often substituted with accented equivalents, or similar-looking numbers (l->1, O->0, e->ê, f->, etc). phPOP3clean will scan for these phrases that match with at least 1 character substituted, but don't match exactly (so an email with the word "viagra" in it will not be deleted, but "víàqrä" will be deleted). Source Code words/phrases are matched on the HTML portion of the message only, without stripping HTML tags. This is used to match known exploits, and for phishing emails where the IP is not constant (so there's no point blacklisting it) but the URL structure is. For Clean and Source Code phrases, the entered phrase can either be plain text, or a Perl-syntax regular expression (if the box is checked). In the latter case, you're responsible for escaping your own regular expression characters. Note for regular expression mode: * Use hex characters for HTML entities in regular expressions, for example "\xA0" instead of "&nbsp;" * Use \s instead of a normal space inside bracketed expressions good: [\sa-z]+ bad: [ a-z]+ ---------------------------------- Administration - "Received" header ---------------------------------- This scans the "Received" header of each email and matches it against blacklisted domains. This filtering technique should generally be avoided but may be useful for some large-volume spammers that are otherwise hard to filter. ---------------------------------- Administration - Whitelist (Email) ---------------------------------- The whitelist bypasses all filtering. The whitelisted email addresses are matched against the "From" and "Return-Path" headers and if either matches the email is excluded from further scanning. ------------------------------------ Administration - Whitelist (Subject) ------------------------------------ The whitelist bypasses all filtering. If the whitelisted phrase is matched anywhere in the email subject the email is excluded from further scanning. ------------------------------------------- Administration - List recently-seen domains ------------------------------------------- This section lists all domains seen in all scanned emails (whether deleted or not) within the last day (configurable as PHPOP3CLEAN_KEEP_RECENT_DOM) and performs an IP lookup on the domains and shows if they're currently blacklisted or not. There is nothing to configure in this section, and it is not generally useful, feel free to ignore it.