CertaDNS
Skip to lesson

The Controls as One System · lesson 3 of 3

What each one leaves open

After this lesson you can

State, for every control, the attack it does not address.

Assumes you have read The dependency graph.

Every control has a boundary, and the boundary is where the next attack lives. Being able to state each one is what makes a posture assessment honest rather than a list of ticks.

Control by control

ControlLeaves open
SPFAnything not using your domain as the envelope sender — which includes most spoofing, since the From header is what a reader sees.
DKIMReplay of a validly signed message, and anything sent by an authorised platform whose account is compromised.
DMARC at rejectLookalike domains, display-name spoofing, compromised mailboxes, compromised suppliers. Four of the five impersonation techniques.
DNSSECAnything the zone owner publishes deliberately. It proves authenticity, never correctness.
CAAIssuance by a CA that does not check it, and issuance for names under a delegation you no longer control.
MTA-STSSenders that do not implement it, everything after the message is accepted, and your own outbound transport.
DANEThe same, plus every receiver whose provider is unsigned.
BIMIEverything. It is a perception aid with no security property of its own.
Registry lockCompromise of your DNS provider, which is a different account entirely.
Domain monitoringSix of the seven impersonation techniques, because they involve no registration.
TakedownEverything that happened before it, and the next domain.

The pattern

Read down the list and the same shape recurs: each control makes one kind of claim checkable, and an attacker moves to a claim nobody is checking. That is not a failure of the controls — it is what a working control does. The value is in how much it costs the attacker to move, and in whether you can see them when they do.

State these before anyone asks

A posture report that lists what is deployed invites the question “so are we safe?”. One that lists what each control leaves open answers it in advance, and it is the difference between a programme that is trusted after an incident and one that is blamed for it.

Last reviewed