Sigma rule testing

Test Sigma on your own history before it pages anyone.

Most detection rules ship without ever seeing your data. In Huntbase, every Sigma rule is compiled onto OCSF, backtested on your history, run in shadow and versioned like code.

The lifecycle

From rule to watcher, with evidence at every stage.

Five stages. A rule moves on when the evidence says it is ready, and goes back when it floods.

  1. Write or import

    Your own Sigma, or SigmaHQ community rules, which arrive switched off.
    Off
  2. Compile

    Native Sigma compiled onto OCSF fields, so one rule runs on every matching source.
    Bound to OCSF
  3. Backtest

    Run it over up to 90 days of your own history and see what it would have fired on.
    Up to 90 days
  4. Shadow

    Evaluated on live data and recorded. It reaches the digest and pages nobody.
    Digest only
  5. Canary, then on

    A staged rollout once the evidence says it is ready. A rule that floods goes back to shadow.
    Staged

The backtest

See what a rule would have fired on last quarter.

Pick a window of up to 90 days and Huntbase runs the rule over your own history. You see how often it would have fired, on which hosts, and a sample of the events.

The fast count is approximate, so Huntbase also runs the rule's exact logic over a sample and tells you when the two disagree.

Watcher · encoded PowerShell · v3
title: Encoded PowerShell command line
logsource: { category: process_creation, product: windows }
detection:
  selection:
    Image|endswith: '\powershell.exe'
    CommandLine|contains: ' -enc '
  condition: selection
tags: [attack.t1059.001]

Compiles · bound to process_activity · process.file.path, process.cmd_line

Would have fired
9 in 90 days
Exact on sample
9 of 9
Hosts
3
Backtest: 1d · 7d · 30d · 90dOffShadowCanaryOn

What you get

Rules you can trust, and the evidence to show why.

  • SigmaHQ, switched off

    Enable the community rules and they land as watchers, off by default. Rules that cannot run on your data are grouped by reason, so the gaps become a to-do list.
  • Backtests that show their working

    Huntbase shows the fast approximate count and the rule's exact logic over a sample, and says so when the two disagree.
  • Versions with diffs

    Every version keeps its author, note and diff, and you can restore any of them. Every firing names the version that fired.
  • Correlation rules

    Sigma event count, value count and temporal ordering rules run on the same correlation engine.
  • Your other SIEMs

    Convert any rule to Splunk SPL, Microsoft Sentinel KQL or Elastic ES|QL for data that stays in those tools.
  • From hunt to watcher

    Scout drafts a watcher from confirmed hunt findings. You backtest it and decide the rollout.

Read next:Detection EngineeringIntelligence in HuntbaseTest detections against real attacksWhat is Sigma?

Questions

Sigma rule testing, answered.

How do I test a Sigma rule before deploying it?

In Huntbase, backtest it over your own history for up to 90 days to see what it would have fired on, then run it in shadow mode, where it is evaluated on live data but only reaches the digest. Turn it on when the evidence says it is ready.

Can I import the SigmaHQ rules?

Yes. Enable the SigmaHQ community rules for your organisation and Huntbase fetches them for you. They arrive as watchers, switched off. Rules that cannot run on your data are grouped by reason, such as a field that is not mapped.

Does Huntbase convert Sigma to other query languages?

Huntbase compiles Sigma natively onto OCSF-normalised data. For data that stays in another SIEM, any rule can also be converted to Splunk SPL, Microsoft Sentinel KQL or Elastic ES|QL.

Are Sigma correlation rules supported?

Yes. Sigma correlation rules (event count, value count and temporal ordering) run on Huntbase's correlation engine.

Backtest your next rule on your own data.

Bring your own Sigma or start from SigmaHQ, and see what it would have fired on before it goes live.

  1. Sign up.

  2. Run a guided hunt on the sample workspace. No connector needed.

  3. Connect your own sources when you're ready.