Setting up the SOC

What works out of the box

You do not start from zero. The moment the platform is running, the SOC is already watching itself: every authorization decision and every tool call the platform makes is normalized into a security event and scanned for trouble. This is the built-in dpf.internal source, and it needs no setup.

On the first scan after install, the SOC also seeds its kernel detection pack — a starter set of rules covering the most common attack techniques (cleared audit logs, disabled cloud logging, MFA removal, new privileged accounts, and so on), each mapped to MITRE ATT&CK. These apply to every customer automatically.

So a brand-new install already has:

Open the console and you will see one connected source and a quiet board. That is the baseline. The rest of setup is about widening what you can see.

Step 1 — Open the console and read the baseline

Go to Operations → Security → SOC Console (/ops/security).

The top row answers four questions:

An empty board is not treated as “all green” — if no source is reporting, the console says so, because silence usually means a broken collector, not safety.

Step 2 — Connect your own sources

External activity reaches the SOC through edge collectors — small agents that run close to your systems, read native logs, and send a normalized summary to the platform. The raw log stays where it was produced; only the summary travels.

The platform ships reference collectors for:

Source (sourceKind) What it watches
windows.security Windows Security event log — logons, account and group changes, audit-log tampering, service and scheduled-task creation.
aws.cloudtrail AWS control-plane activity — logging tampering, IAM and MFA changes, access-key creation.
cef Any product that emits ArcSight CEF over syslog — firewalls, IDS/IPS, and other security appliances.

Broader collector coverage (endpoint/EDR, identity providers, more cloud platforms) expands over time; new sources plug into the same pipeline without changing anything downstream.

To connect a source:

  1. Enroll an edge node. Under Platform → Edge Nodes, create a node for the site or customer you are collecting from. The platform issues a one-time enrollment token. This is the trust boundary — the node’s identity, not the contents of any message, decides which customer the data belongs to.
  2. Run the collector on that node. Point the reference collector at the log source and at your platform’s events endpoint. It authenticates with the node’s token and posts security events.
  3. Confirm it is reporting. Back on the SOC Console, the Connected sources count goes up and the new sourceKind appears. If a source was reporting and then goes quiet, the console flags it as stale rather than hiding it.

Tenancy is set at the boundary, not in the message. Because scope comes from the enrolled node, a collector cannot claim to be a different customer. This is what makes the multi-customer (MSP) model safe.

Step 3 — Add threat intelligence (optional)

The SOC can match activity against known-bad indicators (IPs, domains, file hashes). Load your threat-intelligence indicators and the correlation sweep will raise a high-severity detection whenever an event’s observable matches an active indicator. Indicators carry a validity window, so expired intel stops matching on its own.

Step 4 — Decide your autonomy posture

The SOC’s willingness to act on its own is derived from your organization’s risk posture, set during onboarding — you do not configure a separate security autonomy dial. A cautious posture means more actions wait for human approval; a confident posture lets reversible, low-impact responses move faster. Either way, irreversible or estate-wide actions always need a person.

You can review and adjust the underlying posture in your organization’s onboarding/risk settings; the SOC follows it automatically.

What you should see when setup is done

From here, move on to Operating the SOC.