Spam filter
Two mechanisms decide whether an incoming message is spam:
- Spam scoring examines each message on its own.
- The allow and deny lists overrule that decision for senders you've already decided about.
Sorting accepted mail into folders is a separate job. See Sieve filtering.
The spam score and the threshold
Each incoming message gets a spam score based on its content, the sender's reputation, and the authentication results.
| Score | What happens |
|---|---|
| At or above the Spam Score Threshold | The message is marked as spam and goes to the mailbox's Junk folder instead of the inbox. |
| Far above the threshold | The server refuses the message. |
The threshold defaults to 6. You set it in two places:
- as the default for the whole mail service
- as an override for one mailbox, under the domain user's Email Access Policy section. An empty field means "use the service's default".
A lower value catches more spam, and marks more good mail as spam.
One address that gets a lot of spam is no reason to make the whole domain stricter.
A forwarder refuses suspected spam instead of forwarding it. Forwarding spam to an external destination would damage your domain's reputation there.
For comparison, mail to a group goes to each member's Junk folder as usual, and a mailing list verifies a post before distributing it.
Allow/Deny Lists
The Allow/Deny Lists page holds the sender rules for your whole account: "never mark mail from our payment processor as spam", or "refuse everything from this domain".
Each entry names two things:
- a Sender: a full address such as
[email protected], a domain such asexample.com, or a subdomain wildcard such as.example.com, which matches the domain and everything below it - a Decision: Allow or Deny
When several entries match a message, the most specific wins: a full address over a domain, a domain over a subdomain wildcard. The rules apply to your whole account, before each mailbox's own filtering.
An entry can also carry Allowed Origins: the networks or hostnames the sender's mail must come from. Mail from anywhere else doesn't get the Allow decision.
The sender's SPF, DKIM, and DMARC records already prove the origin.
The field is for the unusual sender whose mail is legitimate but fails authentication, e.g. a misconfigured appliance or an old relay. Pinning the origin lets you allow that sender without also allowing anyone who forges the address.
Spam learning
The filter also learns from how people handle mail.
| Client | How to teach it |
|---|---|
| Webmail | The Mark as spam and Not spam actions. |
| A normal email client | Move the message into the Junk or Spam folder to mark it as spam, or out of that folder back to the inbox to mark it as not spam. |
Learning moves the score slowly, over time. An allow or deny entry is an immediate overrule. The two work well together.