Skip to main content
Version: 0.10

Integrations

An integration is a configured source of asset data. Most are a connection to a system that already knows something about your estate — an EDR, a directory, a cloud subscription — which Velum Edge reads on a schedule and folds into the asset inventory. Two of them, the active and passive scanners, read the network itself instead.

Every integration in the catalogue below has its own page covering what it contributes, what you have to configure, and what it needs permission to read in the source system. This page covers the parts they all share.

How an integration runs​

Integrations do not run in the console. Each one is assigned to a sensor, and the sensor is what makes the outbound calls. A run goes through the same stages regardless of type:

Two consequences are worth internalising:

  • The sensor makes every call, so it needs network reach to the source system — including through a proxy if one stands in the way. Credentials are stored with the integration instance in the console and handed to the sensor with the task, so they are used from inside your network but not kept there. Secret fields (passwords, client secrets, API keys) are write-only: once saved, the console shows them as a locked box you can Replace (or Remove, where the field is optional) but never displays the value again. Only the sensor running the integration receives it.
  • Nothing is written straight to an asset. A run produces change entries, and what happens to them is governed by the integration's action settings described below.

Configuration model​

Every integration has two layers of configuration. The first is the connection itself, and is specific to the type — see each integration's page.

The second layer is a scan template. Its common settings are near-identical everywhere; options specific to one integration are listed on that integration's page.

GroupSettingWhat it does
BasicTemplate name, Template DescriptionLabels the template
AdvancedIgnore of host is offlineHow long a record may go unseen before it stops refreshing the asset. Some integrations label it after what they read, for example identities or logins, and a few do not have it
AdvancedScan TimeoutUpper bound on a single run
ActionsOn New Asset DiscoveredCreate Asset Automatically (default), To Approve, or Ignore
ActionsExisting asset updatedUpdate Asset Automatically (default), To approve, or Ignore
ActionsAsset missingDo nothing (default), To approve, or Delete asset
ActionsOn Scan Complete / On Scan FailedSend email

The setting labels above are spelled as the console shows them.

Start with "To approve" on a new integration

Setting On new asset discovered to To approve for the first few runs lets you see what a source actually asserts before it can create anything. It is much easier than unpicking a few thousand automatically created assets afterwards.

"Delete asset" acts on the source's silence

Asset missing fires when the source stops returning a record — which happens when an asset is decommissioned, but equally when a credential expires or a console-side filter changes. Prefer To approve.

How integration data lands on an asset​

Each integration writes two kinds of field:

  • Prefixed fields, namespaced to the integration — cs_* for CrowdStrike, defender_* for Windows Defender, nessus_* for Nessus, and so on. These are the source's own view, kept verbatim so you can always see who said what.
  • Shared asset fields — names, os, os_vendor, type, private_ip, mac, updated_at and friends. Several integrations write these, which is precisely the point: they are what correlation uses to decide that the CrowdStrike host and the Nessus agent are one machine.

Each type also sets a boolean presence flag — cs_exits, nessus_exits, azure_md_exits — which is the cheapest way to filter for "assets this source knows about", and by inversion, coverage gaps. Older integrations spell the flag _exits and newer ones _exists (tsc_exists, ls_exists); each integration's page gives its own.

A few shared fields follow rules of their own, whichever integration writes them:

  • mac collects every network interface any integration knows for the machine. Each integration adds its MACs to the ones already there instead of replacing them. CrowdStrike, Windows Defender, Elastic agent, Nessus agents, Tenable Security Center, Vectra NDR, Azure VMs and the active scanner leave out placeholder addresses that many machines share, such as an all-zero MAC, the broadcast address or a virtual adapter installed with the same address everywhere, because they would tie unrelated machines together.
  • private_ip is the machine's own address on your network. Addresses of virtual interfaces (container bridges, VPN clients, cluster heartbeat links) are passed over. Integrations whose agents report every interface offer a Virtual interface ranges setting in the scan template for ranges specific to your estate.
  • first_seen is when an integration first saw the machine, and last_seen keeps the most recent sighting across all integrations. A time a source does not report is left empty rather than filled with a placeholder date.

Where two sources disagree on a shared field, some integrations write in do-not-override mode so they enrich an empty field but never clobber a populated one. The asset detail view shows which source asserted each value; see Assets.

Catalogue​

Network discovery​

These two do not read an external system at all — the sensor itself is the source, probing or observing your network directly. They are how you find assets that no agent and no cloud API knows about, so they belong at the top of any rollout plan.

IntegrationContributes
Active scannerHosts, open ports, services and device identity, established by probing
Passive scanner (WIP)Hosts inferred from mirrored traffic, without sending anything (not yet available)

Endpoint detection and response​

IntegrationContributes
CrowdStrikeManaged endpoints
CrowdStrike process events (WIP)Executables seen running on managed endpoints, written onto existing endpoints
CrowdStrike applications (WIP)Business applications and data platforms on servers: installed, in containers and running
Carbon Black (WIP)Managed endpoints
Windows Defender (WIP)Machines onboarded to or known by Defender for Endpoint

Cloud and virtualisation​

IntegrationContributes
Azure managed devices (WIP)Devices registered or joined to Entra ID
Azure VMs (WIP)Virtual machines from Azure subscriptions

Directory and identity​

IntegrationContributes
Active Directory (WIP)Computer objects
Active Directory users (WIP)User identities
LDAP usersUser identities from a generic LDAP directory
Azure users (WIP)User identities from Entra ID
CrowdStrike identities (WIP)User identities from Falcon Identity Protection, one per person across their directory and SSO accounts
LinkShadow users (WIP)User identities from the directories LinkShadow is joined to, and usernames seen only on the network
Atlassian users (WIP)User identities from Atlassian Cloud

Vulnerability and network​

IntegrationContributes
Nessus agentsLinked Nessus agents, with platform, link state and last scan
Tenable Security Center (WIP)Scanned hosts, and every vulnerability finding on them, open and fixed, with its history
Vectra NDR (WIP)Network-detected hosts
LinkShadow assets (WIP)Network-detected hosts, one per machine, with their addresses, directory details and real last activity
LinkShadow DHCP (WIP)Devices from DHCP leases: MAC, name and address, matched to the endpoints you already have
LinkShadow applications (WIP)Applications seen on the network: what servers run, the sites and cloud services people use, and what depends on what

Security monitoring​

IntegrationContributes
QRadar log sources (WIP)Hosts that send logs to QRadar, with their log sources and when each last reported

Elastic​

IntegrationContributes
Elastic agentHosts reporting an Elastic agent
Elastic aggregation query (WIP)Values from an aggregation over your own indices, written onto existing endpoints

Ownership enrichment​

These are consulted only to resolve who owns an asset; they do not create assets. They fill the owner, and the department and criticality where the asset has none.

IntegrationResolves owner from
Elastic asset owner (WIP)Logon events in an Elasticsearch index
CrowdStrike asset owner (WIP)CrowdStrike logons, Identity Protection and Discover

Adding an integration​

  1. Integrations → Add integration.
  2. Pick the type.
  3. Supply credentials and scope, and choose the sensor that will run it.
  4. Save. The first sync is queued.
Use a dedicated service principal

Give each integration its own read-only service account rather than reusing an existing one. It makes revocation and audit straightforward, and it keeps Velum Edge's access visible in the source system's own logs.

Monitoring integration health​

Each integration's detail view shows its last run, the record count and any errors. An integration that fails silently is the most common cause of an inventory that "looks stale" — check here first.

Credentials expire

Cloud and SaaS credentials rotate or lapse. A failing integration stops refreshing its assets but does not remove them, so coverage degrades quietly. Watch for repeated auth failures.

See also​

  • Core concepts — how sources combine into one asset
  • Assets — inspecting which source asserted an attribute
  • Sensors — the component that actually runs each integration