MTAs move email from one host or network to another. These are in contrast to Mail Delivery Agents (MDAs), which move mail within a system (i.e., from an MTA to a local user's mailbox, or from a mailbox to a file or directory). In other words, MTAs are like the mail trucks (and airplanes, trains, etc.) that move mail between post offices; MDAs are like the letter carriers who distribute the mail to their destination mailboxes. Procmail is one popular MDA on Linux systems.
In addition to MTAs and MDAs, there are various kinds of email readers, including POP3 and IMAP clients, for retrieving email from remote mailboxes. These clients are also known as Mail User Agents, or MUAs, of which Mutt and Outlook Express are popular examples. There is no real-world simile for these, unless your letters are handed to you each day by a servant whose sole duty is to check your mailbox now and then! But we're not concerned with MUAs or MDAs, except to mention how they relate to MTAs.
Most MTAs support multiple mail-transfer protocols, either via embedded code or separate executables: nearly all MTAs, for example, support at least UUCP and SMTP. Nonetheless, for the remainder of this chapter, I'll assume you're interested in using your MTA for SMTP transactions, since SMTP is the dominant mail-transfer protocol of the Internet.
No matter what other email protocols you support internally, such as the proprietary protocols in Microsoft Exchange or Lotus Notes, you need at least one SMTP host on your network if you want to exchange mail over the Internet. Such a host, which exchanges mail between the Internet and an internal network, is called an SMTP gateway. An SMTP gateway acts as a liaison between SMTP hosts on the outside and either SMTP or non-SMTP email servers on the inside.
This liaison functionality isn't as important as it once was: the current versions of MS Exchange, Lotus Notes, and many other email-server products that used to lack SMTP support can now communicate via SMTP directly. But there are still reasons to have all inbound (and even outbound) email arrive at a single point, chief among them security.
First, it's much easier to secure a single SMTP gateway from external threats than it is to secure multiple internal email servers. Second, "breaking off" Internet mail from internal mail lets you move Internet mail transactions off the internal network and into a DMZ network. Now your gateway can be isolated from both the Internet and the internal network by a firewall (see Chapter 2).
Therefore, I recommend, even to organizations with only one email server, the addition of an SMTP gateway, even if that server already has SMTP functionality.
But what if your firewall is your FTP server, email server, etc.? Although the use of firewalls for any service hosting is scowled upon by the truly paranoid, this is common practice for very small networks (e.g., home users with broadband Internet connections). In this particular paranoiac's opinion, DNS and SMTP can, if properly configured, offer less exposure for a firewall than services such as HTTP.
For starters, DNS and SMTP potentially involve only indirect contact between untrusted users and the server's filesystem. (I say "potentially" because it's certainly possible, with badly written or poorly configured software, to run extremely insecure DNS and SMTP services.) In addition, many DNS and SMTP servers, e.g., BIND and Postfix, have chroot options and run as unprivileged users. These two features reduce the risk of either service being used to gain root access to the rest of the system if they're compromised in some way.
There are several categories of attacks on SMTP email. The scenario we tend to worry about most is exploitation of bugs in the SMTP server application itself, which may result in a disruption of service or even in the hostile takeover of the underlying operating system. Buffer-overflow attacks are a typical example, such as the one described in CERT Advisory CA-1997-05 (MIME Conversion Buffer Overflow in Sendmail Versions 8.8.3 and 8.8.4 ? see http://www.cert.org/advisories/CA-1997-05.html).
Another danger is abuse of the SMTP server's configuration, that is, using the server in ways not anticipated or desired by its owners. The most widespread form of SMTP abuse is relaying. Spammers and system crackers alike rejoice when they find an SMTP server that blindly accepts mail from external entities for delivery to other external entities.
Such " open relays" can be used to obfuscate the true origin of a message and to forward large quantities of Unsolicited Commercial Email (UCE) and other undesirable email. For example, open SMTP relays were an important attack vector for the "Hybris" worm as described in CERT® Incident Note IN-2001-02 (Open mail relays used to deliver "Hybris Worm," http://www.cert.org/incident_notes/IN-2001-02.html).
Still another security risk in SMTP is that one's MTA will leak user and system information to prospective intruders. Like SMTP abuse, SMTP "intelligence gathering" usually capitalizes on sloppy or incorrect software configuration rather than bugs per se.
The main difference between abuse and probes is intent: somebody who relays UCE through your server probably doesn't care about the server itself or the networks to which it's connected; they care only about whether they can use them for their own purposes. But somebody who probes an SMTP server for usernames, group memberships, or debugging information is almost certainly interested in compromising that SMTP server and the network on which it resides.
Historically, two SMTP commands specified by RFC 2821 (Simple Mail Transfer Protocol, available at ftp://ftp.isi.edu/in-notes/rfc2821.txt) have been prolific leakers of such information: VRFY , which verifies whether a given username is valid on the system and, if so, what the user's full name is; and EXPN , which expands the specified mailing-list name into a list of individual account names.
A third SMTP command, VERB , can be used to put some remote MTAs into "verbose" mode. VERB is an Extended SMTP command and was introduced in RFC 1700 (Assigned Numbers). Since one of the guiding principles in IS security is "never reveal anything to strangers unnecessarily," you should not allow any publicly accessible MTA server to run in verbose mode.
EXPN, VRFY, and VERB are throwbacks to a simpler time when legitimate users wanting such information were far more numerous than mischievous strangers up to no good. Your MTA should be configured either to ignore VRFY and EXPN requests or to falsify its responses to them, and to disregard VERB requests.
Unsolicited Commercial Email (UCE) isn't a security threat in the conventional sense: sending UCE generally isn't illegal, nor is it a direct threat to the integrity or confidentiality of anyone's data. However, if somebody uses your bandwidth and your computing resources (both of which can be costly) to send you something you don't want sent, isn't this actually a kind of theft? I think it is, and many people agree. Rather than being a mere annoyance, UCE is actually a serious threat to network availability, server performance, and bandwidth optimization.
Unfortunately, UCE is difficult to control. Restricting which hosts or networks may use your SMTP gateway as a relay helps prevent that particular abuse, but it doesn't prevent anyone from delivering UCE to your network. Blacklists, such as the Realtime Blackhole List (http://mail-abuse.org/rbl/), that identify and reject email from known sources of UCE can help a great deal, but also tend to result in a certain amount of legitimate mail being rejected, which for some organizations, is unacceptable. Anyhow, blacklists are a somewhat crude way to address UCE.
A much better approach is to use scripts such as SpamAssassin (available at http://www.spamassassin.org) to evaluate each incoming email message against a database of known UCE characteristics. With some fine tuning, such scripts can radically reduce one's UCE load. Depending on the volume of email arriving at your site, however, they can also increase CPU loads on your SMTP gateway.
SMTP exploits, relaying, and abuse, including UCE, are all SMTP problems; they're risks endemic to the SMTP protocol and thus to many SMTP Mail Transfer Agents. But surely there's some proactive security feature in SMTP?
Until recently, there wasn't: SMTP was designed with no security features at all, not even the most rudimentary authentication mechanism. But that changed in 1999 with the introduction of RFC 2554, SMTP Service Extension for Authentication (known more simply as "SMTP AUTH"), which provided the SMTP protocol with a modular authentication framework based on the generic Simple Authentication and Security Layer (SASL) described in RFC 2222.
SMTP AUTH allows your MTA to authenticate prospective clients via one of several authentication schemes. In this way, you can more effectively control such activities as SMTP relaying, and you can also provide SMTP services to remote users, even if their IP address is unpredictable.
It's far from a panacea, and it isn't even supported by all MTAs, but SMTP AUTH is a badly needed improvement to the venerable SMTP protocol. Both MTAs we discuss in this chapter support SMTP AUTH.