Vectra NDR (WIP)
Pulls the hosts Vectra has observed on the network, along with its threat and certainty scoring and the artefacts it used to identify each one.
| Type name | vectra_ndr |
| Creates | Endpoints |
| Runs on | A sensor with network reach to the Vectra appliance |
| Presence flag | vectra_exists |
| Field prefix | vectra_ |
What it contributes
Vectra sees hosts by their traffic, so it finds machines that carry no agent and appear in no directory.
- Identity — host ID, name, LUID, URL, host URL
- Network — IP, previous IPs, last source, active traffic flag
- Scoring — threat score (
t_score), certainty score (c_score), severity - Detections — detection set, detection profile, last detection timestamp, custom model flag
- Criticality — key asset flag, targets-key-asset flag
- Ownership signals — probable owner, privilege level, privilege category
- Workflow — state, assigned to, assigned date, past assignments, notes and note modification metadata, tags, groups
- Host artefacts — MAC, rDNS, DNS, DHCP, Kerberos, NetBIOS, VM UUID, VM machine info, Windows Defender name
It writes the shared fields:
| Shared field | Derived from | Mode |
|---|---|---|
names | The Vectra host name, short form | Merge: added beside other sources' names |
fqdn, domains | The Vectra host name, when it is fully qualified | Merge |
private_ip | The host's current IP | Overwrite |
mac | Every MAC artefact Vectra holds for the host. Placeholder MACs that many machines share are never written | Merge: added beside other sources' MACs |
updated_at | Time of the run that produced the record | Overwrite |
It does not set the endpoint's first_seen or last_seen.
Which hosts are kept
Not every host Vectra reports becomes an asset change:
- The sensor reads only hosts with a detection on or after 1 September 2024.
- A host is kept only when its IP falls inside a subnet already in the inventory, and every inventory subnet that contains it is marked as static. A host in an unknown subnet, or in a subnet not marked static, is dropped and recorded in the task log.
So before the first run, make sure the subnets Vectra watches are in the inventory with their static flag set; otherwise the run completes and creates nothing.
The detection-date cut-off and the static-subnet rule are fixed in the current release. Confirm whether they are to become scan template options, and document them here if so.
How results match existing assets
A host is matched to an existing endpoint by private_ip alone. This is why only
hosts in static subnets are kept: in a DHCP range the same address belongs to
different machines over time. Where more than one endpoint holds the address, the
change entry is marked as needing manual input and lands in the approval queue.
vectra_probable_owner is a free ownership signalVectra infers a likely user from observed authentication traffic. It is weaker than a directory-sourced owner but costs nothing extra, and it covers hosts the ownership integrations cannot see.
vectra_host_artifact_set_mac, _dns, _netbios and _vm_uuid are what let a
traffic-derived host line up with an agent-derived one. Matching does not use them
today (see How results match existing assets),
but the MACs reach the shared mac field, and the artefacts are worth surfacing in
a layout if you rely on Vectra for coverage of unmanaged segments.
Requirements
The integration is outbound-only: it reads the Vectra API on a schedule, so nothing is installed on the appliance, no traffic is mirrored to the platform, and nothing has to reach into the deployment.
Network access
| From | To | Protocol | Port |
|---|---|---|---|
| The sensor running the integration | The Vectra appliance named in URL | HTTPS over TCP | 443, unless the appliance URL names another |
One host: the token is sent as a request header rather than exchanged at a separate token endpoint, so the paged host read is the only call made. The appliance is an internal destination, so the rule belongs in the internal firewall rather than at the perimeter.
If the appliance presents a certificate the sensor host does not trust, install the CA on the host in preference to enabling Ignore SSL — the token is sent on every request.
Access in Vectra
| Need | Why |
|---|---|
| An account that can issue API tokens | The credential is a Vectra API token — see Permissions in Vectra |
| That account holding a read-only role with host access | Vectra tokens carry their owner's permissions, so the role, not the token, decides what the run can read |
| An appliance version exposing API v2.5 | The run pages /api/v2.5/hosts |
In this deployment
| Need | Why |
|---|---|
| A registered, healthy sensor with reach to the appliance | The integration runs on a sensor, on the platform's schedule — see Sensors |
There are no special host requirements beyond a sensor that is checking in — this integration reads Vectra's findings rather than observing traffic itself, so it needs none of the NIC configuration a passive sensor does.
Configuration
| Field | Required | Notes |
|---|---|---|
| Name | Yes | Label for this instance |
| URL | Yes | Vectra appliance base URL |
| Token | Yes | Vectra API token |
| Cookies | No | Extra cookies added to requests, sent as written in a Cookie header |
| Sensor | Yes | Sensor that will run the integration |
| Ignore SSL | Yes | A checkbox: on skips certificate validation |
Token and Cookies are secrets: once saved, the console shows each as a locked box with Replace (and Remove for Cookies), never the value, and sends it only to the sensor.
The run pages /api/v2.5/hosts 500 hosts at a time, filtered to hosts with a
detection on or after 1 September 2024, following the next link until exhausted
and trying each page up to three times. The connection test reads a single host
with the same token.
Permissions in Vectra
Issue an API token for an account with read access to hosts. Vectra tokens carry the permissions of their owning account, so create one under a read-only role.
Confirm which Vectra roles grant read on /api/v2.5/hosts, and note the minimum
appliance version exposing v2.5.
See also
- Sensors: where the integration runs
- Passive scanner (WIP): the platform's own traffic-based discovery, intended to cover the same blind spot, not available in this release
- Integrations overview — the shared configuration model