Skip to content

Custom rules, saved filters, and ignore rules

The built-in detection catalogue covers the common attacks. Real fleets need three more things: rules for what is specific to you, a way to apply one configuration to many hosts, and a way to silence events you have investigated and decided are fine. This page covers all three and the Security, then Detections page where most of it lives.

Security, then Detections is the single configuration surface for what the agents detect and how it alerts. It replaced the older Policies and Rules pages, which now redirect here. Five tabs:

TabHolds
RulesThe policies (access control, file integrity, network security, process monitoring, container security, privilege escalation, cryptojacking) and their rules, with on/off and severity overrides
AdvancedIndividual rules, frameworks (general, SOX, PCI DSS), and per-host overrides including custom rules
ListsAccount-wide lists the rules read: the SSH allowed-source list, authorised container registries
AlertingThe bridge from security events to alerts: on/off, minimum severity (default high), and the escalation policy
ResponseAutomatic SSH brute-force blocking: dispatch, thresholds, and the kill switch that forces dry-run; see SSH login monitoring

Every change is reconciled server-side and pushed to the affected hosts as agent configuration. A policy shows Degraded when a host cannot run all the probes it needs. Details of the Rules and Advanced tabs are in Customizing detection rules.

A custom rule is evaluated on the host by the agent, like a built-in rule. Add one from the per-host override on the Advanced tab.

FieldNotes
Name, descriptionShown on events the rule raises
ConditionThe match expression, in the same form as the built-in rules
SeverityDefault medium
PriorityThe syslog-style priority the event carries; default WARNING
Tags, MITRE tactics and techniques, frameworkOptional classification, carried on every event

Saving links the rule to the host’s override and pushes the probes the rule needs; the agent evaluates the condition locally, so a rule costs nothing on the server side. Rule IDs are prefixed custom- in the event stream. A background sync retries any host that missed the push, every minute, until it confirms.

A second, older kind of custom rule lives at Security, then Custom Rules (the list now redirects to Detections; the add and edit forms remain). These are not evaluated by the agent. They describe a policy in plain terms - allowed users, blocked processes, time windows, hosts or profiles they apply to, and a free-text condition - and that description is given to the AI triage model as context when it judges an event. Use them to tell the triage “backups run as svc-backup between 01:00 and 03:00” so a legitimate pattern stops being flagged as suspicious; use a detection rule when you want a deterministic event.

A saved rule filter is a named configuration - which policies are on, which frameworks, and per-rule severity overrides - that can be applied to hosts. Find them under Security, then Filters (the older Security Profiles link lands here).

Create one, set the policies and overrides, and choose Apply to host. Applying switches the host from “use workspace defaults” to that exact configuration, written into the host’s agent configuration and pushed. Editing the filter later does not re-apply it; apply again to the hosts that should follow it. Typical filters: “web tier” with network rules tightened, “build agents” with process rules relaxed, “PCI hosts” with the PCI DSS framework on.

An ignore rule suppresses events, not configuration. When an event matches, it is dropped before triage and alerting, and the rule’s match counter increments so you can see it working. Find them under Security, then Ignore Rules, or create one from an event’s page with the event’s fields pre-filled.

FieldNotes
Scopehost, process, user, rule, rule_host (default), rule_process, or global
MatchAny of rule name, hostname, agent, process name, process path, command line, user, container, priority
ReasonRequired; future you will want it
ExpiresOptional. Set one for a temporary suppression during a migration; leave empty for a permanent decision
EnabledToggle without deleting

Prefer the narrowest scope that stops the noise: rule_host (this rule on this host) or rule_process (this rule from this binary) rather than rule or global. The source event is linked from the rule so the decision can be revisited.

You want toUse
Detect something the catalogue does notA custom detection rule
Explain a legitimate pattern to AI triageA context rule
Give a group of hosts the same detection configurationA saved rule filter
Stop a specific event that keeps firingAn ignore rule
Lower a built-in rule’s severity everywhereA severity override on the Rules tab
Silence everything on a host during planned workA maintenance window, not an ignore rule