nano SIEM
Hunts

Escalation

Supported in:Self-hosted

Control how hunt findings reach your analysts: the escalation bar, the off, notify, and case modes, the case gate, notifications, and acknowledgement.

Choose how hunt findings reach your analysts. Only the reviewer can escalate a finding. It must have strong evidence, and the hunt's escalation mode decides what happens next.

Before you begin

Make sure the reviewer is on for the hunt. It's on by default for hunts that run on the server, and off by default for hunts that run in nano Desktop. See Turn the reviewer on or off.

How the reviewer decides

The reviewer decides from the evidence before it can read the investigator's reasoning. nano enforces this order and records it on each verdict. The finding page shows whether the verdict was formed before the reasoning was readable.

A verdict counts only if the reviewer cites events it actually read, including at least one of the finding's own events. If it doesn't, nano keeps the finding as a lead and adds a note.

A finding that stayed a lead because the reviewer cited an event it hadn't read

The escalation bar

An escalation goes through only if it meets the escalation bar. The bar depends on the finding's severity and the reviewer's confidence.

SeverityEscalates when
criticalAlways
highThe reviewer's confidence is medium or high (0.5 or more)
mediumThe reviewer's confidence is high (0.8 or more)
low or infoNever

Two more rules apply:

  • Severity can only go up. Each finding starts at the hunt's default severity. The reviewer can raise it, but never lower it.
  • Attempts aren't successes. An attack attempt against an internet-facing system, with no sign that it worked, stays a lead. nano suggests it for the attacker's campaign instead.

If an escalation doesn't meet the bar, the finding stays a lead for an analyst to triage. The finding page shows the reason, for example "critical: always escalates".

The Escalation card on a hunt

Escalation modes

Each hunt has an escalation mode. It decides what happens when a finding is escalated.

ModeWhat happens
offThe finding is labeled escalated and listed. Nothing else happens.
notify (default)The finding appears on the Escalations page, and nano sends a notification.
caseAs notify, and nano also opens a case that needs an analyst's acknowledgement. Available only when the hunt passes the case gate.

No mode takes action on your systems. A hunt never isolates a host, disables an account, closes a case, dismisses a finding, or runs a playbook.

Change the escalation mode

  1. Open the hunt.
  2. On the Escalation card, click off, notify, or case.

If you choose case and the hunt doesn't pass the case gate, nano refuses the change and shows the numbers.

Where escalations land

From v2.0.33, each hunt also chooses where an escalation lands. Set it in the Lands in row of the Escalation card.

ScopeWhat happens
Main (default)The escalation mode applies in the main queue, as above.
TeamThe escalation is held in a team queue. A team member, the hunt owner or an admin promotes it to the main queue or dismisses it. A team queue notifies only the team's own channel.
PrivateThe escalation is held for the hunt owner only.
OffThe finding stays a lead.

Promoting applies the hunt's escalation mode at that moment, so the case gate is checked when a person promotes. Dismissing records a ruling that later runs learn from. The Escalations page has Main, My team and Private tabs.

A SOAR flow can start from an escalation that reaches the main queue. Every action, notification and disposition step on such a flow waits for a human approval.

Escalation floor and duplicates

From v2.0.33, the Escalation floor and duplicates section of the hunt editor has two settings:

  • Escalation floor. The lowest verdict this hunt escalates: low, medium, high or critical (none keeps only the escalation bar). Below the floor, the finding stays a lead. An identical repeat of a lead the floor refused is stored as a duplicate.
  • Duplicate window. 0 to 90 days. A new finding with the same shape as an open finding from this hunt in the window (an unreviewed lead, or an escalation nobody has dismissed or closed) is stored as a duplicate and isn't reviewed again. A worse repeat, such as a higher severity or a success after failures, is never a duplicate. 0 turns the gate off, and blank follows the install default (7 days, set on the Runners page). Hunts that existed before the v2.0.33 upgrade have the gate off.

Users limited to some sources can't change a hunt's escalation mode, reviewer switch, default severity, escalation floor, duplicate window or charter.

The case gate

In case mode, the reviewer can open cases. nano allows this only when the reviewer has a proven track record on the hunt:

  • It must be right at least 90% of the time, over at least 20 escalations that analysts have resolved in the last 90 days.
  • nano judges it conservatively. A small sample counts for less. For example, 18 right out of 20 is 90%, but nano counts it as about 70% until there are more results.
  • The record belongs to the current model. If you assign the reviewer a different model, its record starts again.
  • The current model needs a recent evaluation. nano replays past findings to the reviewer model and compares its answers with your analysts' rulings. Evaluations cost a few cents.

An escalation counts as right when an analyst confirms it, or when its case closes as a true positive. It counts as wrong when an analyst marks it wrong, dismisses it as benign or a false positive, or closes its case that way. Findings ruled duplicate or not actionable don't count.

How long it takes to unlock case mode

Because the gate is conservative, case mode usually needs a track record of about 40 or more resolved escalations. With fewer, one mistake can keep it locked. For example, with 18 resolved escalations, a single wrong one keeps case mode locked.

Until then, use notify mode. Resolve escalations as they come in. Every acknowledgement and every case you close counts toward the track record.

When nano checks the gate

nano checks the gate twice: when you choose case, and again at every escalation. If the reviewer's accuracy drops below the bar, nano sends that escalation as notify instead and records why.

A case that the reviewer opens has no user as its creator. It records the verdict, the confidence, and the reviewer's accuracy at the time. nano writes the audit event hunt.reviewer_escalated_case.

Acknowledge an escalation

  1. Click Hunts, then Escalations.
  2. Open the finding and read the reviewer's reasoning and evidence.
  3. Click Acknowledge.
  4. Choose Confirmed (true positive) or Wrong (false positive).
  5. If the escalation opened a case, close the case from the case page. Acknowledging doesn't close it.

Escalations that need acknowledgement

Each acknowledgement records a ruling. Closing a finding's case also records a ruling.

Send escalations to Slack, Teams, PagerDuty, or email

Escalations use your existing notification channels.

  1. Click Settings, then Notifications.
  2. Add a channel or edit one.
  3. Under event types, select Hunt escalations.
  4. Set the channel's minimum severity. Hunt severity info matches informational.
  5. Save the channel.

nano sends a notification in notify and case modes. It includes the hunt, the finding, the host, user, or IP, the severity, the confidence, the reviewer's model, a link to the case if there is one, and a short reason that nano writes. It never includes AI-written text or event content, so an attacker can't put text into your chat or paging tools through it.

How analyst decisions feed back

Each acknowledgement, dismissal, and case closure becomes a ruling. Rulings:

  • teach later runs which patterns your analysts consider harmless;
  • measure the reviewer's accuracy for the case gate;
  • score evaluations of the reviewer model.

Turn the reviewer on or off

  1. Open the hunt.
  2. On the Escalation card, select or clear Review this hunt's findings with the server AI reviewer.

Without the reviewer, nothing escalates and leads wait for an analyst. For hunts that run in nano Desktop, turning the reviewer on sends the hunt's evidence to the server's AI provider. The server reviews those leads 10 minutes after the sweep ends.

What's next

On this page