Suppose example.com wants to publish SPF. It expects MTAs everywhere to read its SPF record and use it to reject forgery attempts. It hopes SPF reduces the volume of joe-job bounces and bogus abuse reports. So it adds the following line to its zone file:
example.com. IN TXT "v=spf1 a mx ptr -all"
The v=spf1 version string identifies this as an SPF record. The -all means reject all mail by default. Domains that don't send any mail, such as altavista.com, can get by with simply v=spf1 -all. But if the domain does send mail, it declares mechanisms that describe how legitimate mail should look. Mechanisms go in the middle, before -all. The first mechanism to match provides a result for the SPF query. -all always matches and so belongs at the end.
A: the A mechanism means the IP address of example.com is permitted to send mail from example.com. If you want to say the IP address of some-other.com is permitted, you can say a:some-other.com. You can use as many A mechanisms as you want.
MX: the MX mechanism means the MX servers for example.com all are permitted to send mail from example.com. If you want to say the MX servers for some-other.com are permitted, you can say mx:some-other.com. You can use as many MX mechanisms as you want.
PTR: the PTR mechanism says if a host has a PTR record that ends in example.com, it is permitted to send mail from example.com. This would be a good choice for Yahoo, whose mail server names all end in yahoo.com. It would be a bad choice for a broadband provider like Comcast. If you want to say servers whose names end in some-other.com are permitted to send mail from example.com, you can say ptr:some-other.com. You can use as many PTR mechanisms as you want.
IP4: to say the class C network of 192.0.2.0 is permitted to send mail from example.com, you would write ip4:192.0.2.0/24.
Mechanisms are interpreted left-to-right. Using v=spf1 a mx ptr -all first would check whether the connecting client was found in the A record for the domain or, failing that, in its list of MX servers. Then the MTA would check to see whether the hostname of the client matched the domain. If none of the mechanisms matched, -all would be evaluated, the result would be fail and the MTA would be justified in rejecting the mail.
A, MX, PTR and IP4 are enough for the overwhelming majority of domains. The setup wizard at spf.pobox.com/wizard.html can help you configure SPF for your domain. But if your situation is complex, you can use the mechanisms described in the “Advanced SPF” sidebar.
SPF has a number of built-in mechanisms. The basic ones let you designate the hosts that send mail from your domain. This works well for almost all domains out there, because each domain's mail comes only from a small set of hosts. But if mail from your domain is distinguished in some other way, say you always sign it with S/MIME, instead of typing a or mx you can type smime.
Using designated sender mechanisms (A, MX, PTR and IP4) is one possible approach to sender authentication. New sender authentication methods are being developed. SPF is extensible, though, so it can work gracefully with them. SPF plugins that understand future extension mechanisms will be able to interpret them correctly. SPF plugins that don't understand those mechanisms will return unknown, and your domain will be treated as though it did not have an SPF record at all.
Today, spammers forge domain names. Tomorrow, they might forge hostnames. They might try to joe-job your laptop by making up email@example.com. It's a good idea to protect your subdomains as well. You should start with your MX servers and move on to other hosts with A records. Here's why.
Bounce messages are sent with MAIL FROM: <>. The null sender address ensures that bounces don't themselves bounce and create a loop. When SPF sees the null sender address, it falls back to the hostname given in the HELO command. When your MTA sends a bounce message, it announces its hostname in the HELO command it sends. If that hostname has an SPF A mechanism listed, the message passes. So SPF prevents HELO forgery as well.
One Click, Universal Protection: Implementing Centralized Security Policies on Linux Systems
Join editor Bill Childers and Bit9's Paul Riegle on April 27 at 12pm Central to learn how to keep your Linux systems secure.
Free to Linux Journal readers.Register Now!
- Cluetrain at Fifteen
- Getting Good Vibrations with Linux
- Embedding Python in Your C Programs
- [<Megashare>] Watch Mrs Brown's Boys Movie Online Full Movie HD 2014
- New Products
- Customizing Vim
- Memory Ordering in Modern Microprocessors, Part I
- Security Hardening with Ansible
- Tech Tip: Really Simple HTTP Server with Python
- RSS Feeds