OCSF on Axiom
Your Axiom datasets keep events in whatever shape your shipper wrote them:
raw Sysmon, Windows Security, or your own field names. Huntbase can read them
as OCSF, the same normalised shape it uses for
its own data lake. Then one query over a normalised surface, such as
hb_process_activity, returns rows from the Huntbase data lake and from your
Axiom datasets together, with the same columns.
Nothing is copied or rewritten in Axiom. Huntbase turns each normalised query into APL for the dataset and reads it in place.
To make that work, each dataset needs a format: what kind of events it holds, so Huntbase knows how to map its fields.
Dataset formats
Open the Axiom connection's page. The Dataset formats card lists each dataset with its format:
| Chip | What it means |
|---|---|
| A format, such as Windows Security or Sysmon | Huntbase picked it from the dataset's fields. The line under the dataset says why. |
| The same, set by a person | Someone on your team picked it. Schema syncs keep it. |
| Suggested: … | Huntbase thinks it knows the format, but isn't sure enough to use it. Check it and save it. |
| Not mapped | Huntbase found no known format, or someone turned mapping off. Huntbase searches the dataset only in APL. |
Huntbase looks at the dataset's fields when it syncs the schema, and samples which event channels it holds, so a dataset with both Windows Security and Sysmon events gets both formats.
A dataset's format is separate from a field mapping (Apply a field mapping). The format lets normalised queries read the dataset as OCSF. A field mapping translates a watcher's Sigma field names when you deploy it to the dataset.
The formats you can pick are:
| Format | For |
|---|---|
| Sysmon | Raw Sysmon events |
| Windows Security | The Windows Security event log |
| Windows System | The Windows System event log |
| Already OCSF | Events your pipeline already writes as OCSF |
| Custom mapping | Anything else: you map the fields yourself |
Change a dataset's format
- On the dataset's line, click Change (or Map fields when it isn't mapped).
- Pick the format. If the dataset holds more than one kind of event, tick the others under This dataset also holds.
- Optionally, add Field overrides for fields the format doesn't know. Each
row maps a Source field to an OCSF field, such as
process.cmd_line. - Check the preview (below), then click Save.
Pick Not mapped to turn mapping off for a dataset. Huntbase won't pick a format for it again by itself.
You need permission to update the connection to change a format.
Map a custom dataset
For a dataset in your own shape, pick Custom mapping:
- Under Kinds of events in this dataset, pick what it holds: Process activity, Network activity, DNS activity, Sign-ins, File activity or Registry activity.
- Map its fields: a Source field from the dataset, and the OCSF field it fills. Click Add field for each one.
- Preview, then Save.
Map the fields your queries and watchers use, at least the time, host and user. A column you don't map stays empty in the results rather than guessed.
Preview before you save
Preview rows runs a small query over the last 24 hours of the dataset, for the kind of events you pick, and shows a few rows as OCSF. It uses the mapping you are editing, saved or not. Expand The APL Huntbase runs to see the query. If no rows come back, the dataset may simply have none of those events in the last day.
The preview runs one small query on Axiom, which uses a little query compute.
Query normalised surfaces
Once a dataset has a format, a KQL (Huntbase Lake) query over a normalised surface reads it too:
hb_process_activity
| where process_cmd_line has "powershell"
| take 20
- The rows from Axiom carry the connection's name in the store column, so you can tell them apart and narrow to them.
- The run's time range applies to Axiom too. A run without one reads the last 7 days of the dataset.
- The results panel lists the Axiom connection with the other stores it searched. If Axiom fails, the rest of the query still runs, and the connection is listed as excluded with the reason.
- Queries with
joinorletdon't read Axiom datasets.
The same applies wherever Huntbase runs a normalised query: hunt steps, Scout, and Check against history on a watcher, which adds the matches per day from your Axiom datasets.
Each normalised query that reads Axiom runs an APL query there, and uses Axiom query compute. See Keep query compute low.