It seems what ARC needs an A/S for all the people who tried ARC and found
it useless or harmful. Instead of trying to find how to use ARC, we need
to return to the original question of how to detect and block malicious
impersonation. That effort requires understanding where malicious
impersonation happens.
Essentially all malicious impersonation come from clients of bare-metal
hosting services, when those hosting services do not restrict outbound
SMTP. They use the same impersonated email address for both SMTP and
>From addresses. The server names and IP address point responsibility to
the hosting service. The attacker can be completely invisible. The
available remedies are to block messages that have alignment without SPF
Pass, and to send abuse reports to the hosting service. This type of
attack dominates dual-address impersonation by a factor of perhaps 1000 to
1.
This also tells us where to expect malicious impersonation hidden in
messages with two different addresses. It will come from forwarders that
(a) receive malicious attacks from these sources, (b) forward them rather
than blocking them. The risk from this class of traffic is a function of
how much forwarded mail an evaluator organization accepts. Since
forwarding services relay messages from all sources, they are the most
likely source of malicious-but-forwarded mail. By contrast, mailing lists
with a limited set of allowed posters will be difficult to attack and
therefore unlikely to be a source of malicious-but-forwarded messages. An
attack via these lists requires knowing a valid subscriber, not being
blocked by mailing list authentication policy, and being a list with a
useful set of potential victims.
So if one malicious attack comes from forwarding, it is probably another
1000 to 1 ratio between malicious-but-forwarded mail received from
forwarding services vs. malicious but forwarded mail received from mailing
lists.
After an evaluator organization is properly detecting and blocking all of
the malicious impersonations received directly it can begin to consider how
to defend against malicious-but-forwarded mail. There are limited choices:
- Distrust the forwarder: Detect the forwarding point, and assess the
reputation of the forwarder. After concluding that the forwarder is
unacceptable, and refuse all of its forwarded traffic (and probably any
direct traffic as well). Common reasons for refusal are (a) intended
recipients never requested the forwarded traffic, and (b) the forwarding
service is too risky because of poor filtering practices, or (c) the
forwarding point is untrusted for other reasons, such as geo-blocking.
- Trust the forwarder: Detect the forward and assess the reputation of
the forwarder. After concluding that the forwarder is acceptable, choose
to accept the forwarded mail stream without attempting authentication of
the From header address. Unwanted messages can still be blocked based on
From address reputation and content filtering. Malicious impersonation is
a risk, but a risk they have chosen to accept.
- Allow the forwarder, but don't trust his From address filtering.
Instead, look behind the forwarder to detect pre-forwarding value of the
SMTP Mail From address and IP address. Particularly look for messages
that had SPF alignment without SPF Pass prior to the forwarding event.
This becomes possible if: (a) the forwarder preserves the original SMTP
address, (b) the forwarder uses SRS encoding to reveal the original SMTP
address, or (c) the forwarder provides authentication data using
Received-SPF, Authentication-Results, ARC-Authentication-Results, or
Received fields with an (Envelope-From <user@domain>) clause.
What does not make sense:
- Accepting most messages from a mailing list, then blocking mailing
list posts from domains with DMARC p=reject because of changes applied by
the list, then trying to fix the first mistake by trying to use ARC.
- Ignoring malicious impersonation threats hidden in unsigned messages
from a forwarding service.
- Ignoring malicious impersonation from threats hidden in direct
messages that lack DMARC p=reject policies.
If you are solving the impersonation problem, rather than the "mailing
list problem", I do not see how ARC can fail to help.
Doug Foster
_______________________________________________
dmarc mailing list -- [email protected]
To unsubscribe send an email to [email protected]