Fighting Spam at AOL: Lessons Learned and Issues Raised

Presentation Transcript

Carl Hutzler

Director of Anti-Spam Operations

America Online, Inc.

12/9/2005

Agenda

Email Identity Technologies
Email Forwarding
Email Service Provider Best Practices

What do Email Identity Technologies Do?

They provide some assurances that a domain is being used with permission

Citibank can control the use of their domain, but cit1bank.com will still be abused
Bounces can be analyzed to see if they are legitimate
Information can be analyzed on the responsible domain owners and their reputation/accreditation

But remember, email identity technologies do not stop spammers!

They only force spammers into other behaviors, many of which are better for enforcement and controls.
But without message providers doing their part to use these technologies wisely, we will be no better off.

AOL is a Crystal Ball

Bulk Mailers on AOL’s whitelist comprise 30-50% of our daily email volume but only 5-10% of complaints.
>80% of AOL’s spam problem comes from other provider’s main outbound MTAs and compromised web servers (CGI scripts)
AOL began seeing this shift in Sept 2003
The rest of the internet is beginning to see this now…

“We're the biggestspammer on the Internet," network engineer Sean Lutner, Comcast - source CNET.com, May 24, 2004

Report from 9/14/2004

188841 hotmail.com

64543 x-mailer.co.uk

62757 shawcable.com

46312 concentric.net

32259 cnchost.com

32022 zero.ou.edu

23557 mail.atl.earthlink.net

22837 grp.scd.yahoo.com

21005 ucla.edu

17676 oemgrp.com

16849 mail.cornell.edu

16260 dejazzd.com

15764 mta01.tie.cl

15659 mrf.mail.rcn.net

14343 urbanhomesecurity.com

14280 mail.pas.earthlink.net

14246 smtp.nextra.cz

13646 mail.yahoo.com

Note1: Greyed domains have very low spam penetration due to very large number of emails sent which counters the total complaint statistic.

Note2: Italic domains were whitelisted and subsequently blocked for spamming.

All spam will eventually come from Email Message Provider Networks

For example: AOL, BlackLists, and other organizations are getting really fast at blocking zombie machines
BUT…
The machines do not get un-infected
No SMTP AUTH

Most ISPs “trust” internal networks

No Outbound Spam controls
No Rate controls
Results?
ISP mail servers act as forwarding MTAs for a network of open relay Zombie machines

MyDoom’d ZOMBIE PC on DSL.NET

BLOCK

outbound1.dsl.net

mx.aol.com

Hacker/Spammer

Will SenderID, SPF, DomainKeys, etc stop spam?

Simple answer, NO. Complex answer, NO.
Why?

Most AOL spam obeys sender identity technologies TODAY!
Spammers send through the local MTA and use the local ISP’s domain as the FROM/Sender

Identity Technologies can allow blacklists/whitelists to work from DOMAINs instead of IP addresses

Good from a not blocking innocents by IP address standpoint
Reputation/Accreditation systems will be key to success of Email Identity technologies
Without SMTP Authentication, we are only validating the DOMAIN and not the USER portion of the address (user@domain.com)

Bottom Line: If ISPs don’t get smart soon and control the sources of spam on their networks, the reputation for their domain (e.g., comcast.net) will be so poor that they will not have connectivity to other ISPs

Email Forwarding

Forwarding Spam to AOL Customers

AOL can only trust the IP address of the client MTA that connects to an AOL server

No other headers can be trusted as they are all forgeable
This is why internet whitelist/blacklists are all done by IP address.

AOL has no way to no that a message is simply a forwarded email

Does this even matter?

So what happens when a University FORWARDS Spam?

Generally, if AOL gets enough complaints from our members, we block or temp fail the IP address
Is this the members fault?

No, as there is nothing in the email that shows it is from their forwarded account
AOL members do not read headers, nor should they be expected to.

Possible Solutions?

Dedicate an IP address to handle forwarded mail and tell AOL about it.
Do better spam filtering inbound to your network.
Spam filter the outbound traffic and insert a spamassassin x-header that identifies a message as spam. AOL will spam folder it.
Change the headers of forwarded mail to identify the situation to final recipient.

From: ForwardedEmail@university.com
Subject: [FORWARD] Original Subject
ReplyTo: originalsender@originaldomain.com

Bottom Line: Forwarding spam to someone’s inbox innocently or intentionally still creates a bad experience for the final recipient. Port25 is your responsibility.

Mail Service Provider Best Practices

Message Provider Code of Conduct:Take Responsibility for outbound Port 25

ISPs must take full responsibility for all traffic/messages emanating from their network on port25.

Port25 traffic is always Unauthenticated traffic and as such must be accepted by server MTAs.
Abuse issues are always the responsibility of the sending/client MTA

How does a Message Provider like AOL control outbound port25 traffic?

Hijack all direct port25 connections from dynamic IP space to other ISP mail servers and process it for viruses/spam.

Other providers block port 25
Still others use a mail proxy to detect SMTP authentication credentials and only allow authenticated SMTP traffic on port25
Some simply rate limit how much a single IP can send if their IP space is rather static or they can tie an IP to a customer account

Rate limit all customers through outbound, authenticated MTAs. Rate limits per hour and per day work well.
Monitor complaints about customers via the SCOMP Feedback Loop system
URL blocking for known spammer URLs
Secure accounts that are spamming - thousands daily

Summary: What technologies will stop spam?

ISPs and Network Providers “waking up” and working together to cut off the spammer’s oxygen supply:

Spammers need connectivity
Spammers need large numbers of high throughput IP addresses

So what is the formula for success?

ISPs should monitor their networks for sources of spam LEAVING their network

Port25 is always the responsibility of the originating ISP
Shift some of the resources from inbound filtering to OUTBOUND Controls

Enforce strong authentication to authorize use of an ISPs MTAs
Monitor customer sending patterns like a credit company monitors “fraudulent charges”
Monitor/Sign-up to receive complaints from AOL and other sources (spamcop, abuse@, etc)
Remove sources of spam within minutes (Zombie machines, insecure CGI scripts, bad customers, etc)