Alerts and Notifications
Alert types, severity thresholds, digests, and how duplicates are suppressed.
Updated
An alerting system fails in two ways: it misses the thing, or it reports the thing so often that you stop reading. This page describes what SpoofKiller alerts on, how severe it considers each event, and the several mechanisms that keep the second failure from causing the first.
Findings and alerts
Two words are used precisely here.
- A finding is a standing record of a problem on a domain: “SPF exceeds ten lookups,” “potential unauthorized sender at 203.0.113.9.” It has a severity, a status (open, acknowledged, or resolved), a first-seen and last-seen time, and an occurrence count.
- An alert is a notification that something happened: a finding was created, came back, or got worse, or an event occurred that has no standing state, such as a DNS record changing. Alerts appear in the in-app feed and, subject to plan and settings, are sent by email.
Findings and alerts exist only for verified domains.
Alert types
There are 20 alert types. Every one is recorded in the app on every plan. The last column is the plan needed for the alert to be emailed; a client workspace uses its agency’s plan.
| Alert | What triggers it | Email delivery |
|---|---|---|
DMARC policy weakeneddmarc_policy_weakened | A scheduled scan finds the policy moved to a weaker setting (for example reject to quarantine, or quarantine to none): critical. A lowered pct value: high. | Starter and above |
DMARC record deleteddmarc_record_deleted | The DMARC record that was there at the last scan is gone: critical. Also raised as a finding while a verified domain has no DMARC record. | Starter and above |
DMARC record changeddmarc_record_changed | Any other edit to the DMARC record: medium, or high if the rua reporting tag was removed. Also covers a record with configuration errors, such as two DMARC records or no policy. | Pro and above |
SPF record changedspf_record_changed | The SPF record was edited (medium) or removed (high). | Pro and above |
SPF configuration errorspf_configuration_error | A scan finds an SPF error that makes the record fail: no record, more than one record, more than ten DNS lookups, an include loop or missing include, void lookups, a syntax error, or +all. | Starter and above |
DKIM record changeddkim_record_changed | A monitored DKIM key changed (medium) or disappeared (high), or a scan finds a missing, weak, or malformed key or a dangling CNAME at a monitored selector. | Pro and above |
Authentication failure spikeauth_failure_spike | DMARC failures for a reporting day reach at least 50 and at least three times the average of the previous 14 days (and at least 40 above it), given five or more days of baseline. High if ten times the baseline or if half the day's mail failed; otherwise medium. | Starter and above |
Newly observed sending IPnew_sending_source | A source not seen before sends mail as the domain, once the domain has at least seven days of report history, and the source was not flagged as a potential unauthorized sender. Low. | Pro and above |
Potential unauthorized senderunauthorized_sender | A source that is not Approved or Ignored has mail failing both SPF and DKIM alignment, and the impersonation confidence is 40 or more. Low, medium, or high by confidence and volume. | Starter and above |
Approved sender failing authenticationsender_auth_failing | A source you approved (high), or one identified as a known sending service (medium), fails DMARC on more than 20% of at least 20 messages in a report, with no forwarding override noted. | Starter and above |
Authentication domain changedauth_domain_changed | A source known for more than 14 days starts passing DKIM with a signing domain it has never used before. Low. | Pro and above |
New blacklist detectionblacklist_listed | The daily check finds the domain, or the IP address of an approved or self-hosted sender, on a monitored blocklist. High. Resolved automatically when the listing clears. | Pro and above |
Potential lookalike domain discoveredlookalike_domain | The daily check finds a registered domain resembling yours with a risk score of 55 or more, or a known lookalike starts accepting mail. High at a risk score of 80 or more; otherwise medium. | Pro and above |
MX record changedmx_record_changed | The MX records changed (high) or were removed (critical). | Pro and above |
Nameservers changedns_record_changed | The domain's nameservers changed. High. | Pro and above |
DNSSEC problemdnssec_problem | The DS record changed (medium) or was removed (high), or a scan finds DNSSEC validation failing. | Pro and above |
Domain expirationdomain_expiration | Registry data shows the registration expiring within 30 days (medium), 14 days (high), or 7 days (critical). Not raised when the registry publishes no expiration date. | Pro and above |
Ingestion failureingestion_failure | A report file was not accepted: it exceeds the size limit, or the plan's monthly report allowance is used up. Medium. | All plans |
Missing DMARC reportsmissing_reports | A domain whose reporting address is confirmed had been receiving reports and has received none for four days. Medium. Repeated at most once a week. | All plans |
Report processing failurereport_processing_failure | A received file could not be parsed as a DMARC aggregate report. Low. The original file is kept. | All plans |
The key under each name is the value of type in the API. The checks behind the reputation alerts run only on plans that include them, so on other plans those alerts are not raised at all; see pricing for what each plan includes. The sender alerts are explained in detail in Sending Sources, and the DNS ones in domain security monitoring.
A DNS change that strengthens things is logged, not alerted
Severity
Every finding and alert carries one of five severities.
| Severity | Used for | Examples |
|---|---|---|
| Critical | Protection or mail flow is removed outright | DMARC record deleted; policy weakened; MX records removed; registration expires within 7 days |
| High | Likely to cost delivery or to indicate abuse | Approved sender failing; MX or nameservers changed; SPF record removed; blocklist listing; spoofing confidence of 75 or more with at least 50 failing messages |
| Medium | Needs a look soon | SPF or DMARC record edited; failure spike; spoofing confidence of 60 or more; reports missing; registration expires within 30 days |
| Low | Worth knowing | New sending source; signing domain changed; unfamiliar failing source at lower confidence; a report that could not be parsed |
| Info | Recorded for the history | A record published for the first time; a policy strengthened |
For configuration findings, severity comes from the scan check that found the problem. For sender findings it follows the confidence thresholds: an unfamiliar failing source is flagged at a confidence of 40, called potential spoofing at 60 (medium), and rated high at 75 with 50 or more failing messages. A failure of DMARC is not proof of spoofing, so these alerts state their confidence and their reasons for doubt, and the email says plainly when the cause is uncertain.
Organization settings
Owners and Administrators set, for each alert type:
- Enabled. A disabled type creates no alerts, in the app or by email. The underlying finding is still recorded.
- Minimum severity. Events below the threshold create no alert. Setting a type to High means you hear about it only at high or critical.
- Email. Turn email off for a type and its alerts appear in the app only.
- Frequency. Immediate, daily digest, or weekly digest.
Until a type is changed, it is enabled, emailed, and immediate.
Immediate delivery and digests
An immediate alert is emailed when it is created, one message per alert, with the description, the evidence rows, the recommendation, and a link to the finding. A type set to daily or weekly is held instead and sent as one digest listing each pending alert’s title, severity, and domain. Either way the alert is in the app from the moment it is created.
A sensible arrangement is to leave critical matters immediate and move chatty, low-stakes types, such as newly observed sending IPs on a domain with many senders, to a digest.
Separately from alerts, members can receive a weekly digest and a monthly summary of each organization’s domains. Those are reports, not alerts, and each member opts in or out for themselves.
Pausing alerts
An Owner or Administrator can pause alert email for the whole organization for a set time, up to 30 days, and resume early. While paused, alerts are still created and visible in the app; only the email is suppressed. Alerts raised during a pause are not emailed afterward, so read the feed when you come back. Pausing and resuming are recorded in the audit log.
Deduplication
Three mechanisms keep one problem from becoming forty emails.
- One finding per issue. Each finding has a dedupe key that identifies the underlying issue, for example the alert type plus the source IP, or the blocklist plus the listed address. When a check sees the same issue again, it updates that one record: last-seen time, occurrence count, and current details. It does not create a second finding.
- Alerts only on change. For an existing finding, a new alert is raised only if the finding had been resolved and has come back, or its severity has increased. Seeing the same open problem again is not news.
- A 24-hour suppression window. Whatever the cause, a second alert for the same issue is suppressed if one was created in the last 24 hours. Two kinds of event use a longer window: the same DNS record change is not re-alerted for 72 hours, and a missing-reports alert for a domain repeats at most once in 7 days.
Duplicate report files are handled before any of this: a report received twice is processed once, so it cannot double a count or re-trigger detection. See Sending DMARC Reports to SpoofKiller.
Acknowledge and resolve
Analysts, Administrators, and Owners can change a finding’s status.
| Status | Meaning | Effect |
|---|---|---|
| Open | Nobody has dealt with it | Counts as outstanding everywhere, including enforcement readiness |
| Acknowledged | Someone has seen it and it is in hand | Records who acknowledged it. Still counts as outstanding; still updates quietly if seen again |
| Resolved | The problem is fixed or dismissed | No longer outstanding. If the same issue is detected again, the finding reopens and a new alert is sent |
Some findings resolve themselves when the condition clears:
- configuration findings, when a later scan no longer finds the error;
- blocklist findings, when the listing is gone;
- domain-expiration findings, when the registration is renewed;
- unfamiliar-source findings, when you mark the source Approved or Ignored;
- lookalike findings, when you dismiss the lookalike.
Open and acknowledged sender findings hold enforcement readiness at yellow, which is one more reason to close them out rather than let them pile up. Alerts in the feed can be marked read by any member; that is a reading aid and changes nothing else.
Member preferences and unsubscribe
Alert email goes to the direct members of the organization the domain belongs to. For a client workspace that means the people invited to that workspace; agency staff who reach it through inherited access see its alerts in the app but are emailed only if they are also members of it. Two further conditions apply:
- the member’s email address has been confirmed, so that security data is never sent to an address nobody has claimed; and
- the member has alert email turned on for that organization.
Each member has three switches per organization, all on by default: alert email, weekly digest, and monthly summary. They are changed under notification settings, or with the unsubscribe link at the foot of every alert and digest email, which works without signing in. Unsubscribing affects only that member and that organization.
Channels
| Channel | Status |
|---|---|
| In-app alert feed | Available on every plan |
| Available; which alert types are emailed depends on the plan | |
| Slack, Microsoft Teams, webhooks, SMS | Planned. Not available yet |
Until the planned channels exist, GET /api/v1/alerts returns the same feed as JSON on the Pro and Agency plans, and the domain health check remains the quickest way to confirm by hand that an alert’s condition has cleared.
