Mechanisms and Experiences Securing Mail Relay
Chip Rosenthal
<
[email protected]
>
Disclaimers
No source code
Your mileage may vary
The Laser Link Environment
Limited to consumer services
Distributed mail system
Combines open source and proprietary software products
Our Goals
Secured relays
Appropriate to our environment
Distributed implementation
Highly scalable
Happy customers
No special protocols
No special procedures
Allow roaming
Unauthorized Third-Party Relay
IMC reports 83% relays closed. [
http://www.imc.org/ube-relay.html
]
Several mechanisms available
General anti-relay algorithm:
Inspect email address of recipient.
Inspect IP address of sender.
Do
not
use email address of sender - subject to spoofing
Static Anti-Relay
Compare sender IP address to static list
Reference:
http://mail-abuse.org/tsi/
Advantages:
Efficient
Widely supported
Disadvantages:
Leased POPs
Roaming
Laser Link couldn't abide the disadvantages
POP-before-SMTP Authorization
Record IP address after successful POP authentication
Grant that host a ticket to relay
Reference:
http://spam.abuse.net/tools/smPbS.html
Advantages:
No unusual protocol support required
Supports roaming
Disadvantages:
Little support available in existing servers
Need to communicate state between POP and SMTP
Not always transparent to user
Laser Link thought this would be our solution
Our users hated it - particularly those using MS Outlook
Escalating Credentials
Curently deployed in Laser Link network
Amalgamation of two mechanisms:
POP-before-SMTP
Rate limiting
Count mail from originating IP address
Track as messages/hour
At 3 mssgs/hr - require POP-before-SMTP
At 80 mssgs/hr - reject with 450 error
Messages:
550 Relaying denied - example.net customers must check mail before trying to send mail
450 Quota on sending mail exceeded - please try again later
450 Cannot access relay authorization database right now
Advantages:
All advantages of POP-before-SMTP
Eliminated user confusion
Beyond anti-relay - makes us unattractive to spammers
Disadvantages:
Complex implementation
Will users accept upper limit?
False positive to conventional relay testing
Email Abuse Roundtable / October 6, 1999