SOC Mastermind · System3

Closes the alerts you already know. Leaves the doubtful ones to you.

SOC Mastermind is a SOC assistant that runs next to your SOAR. It closes recurring false positives with evidence and a written reason, and routes every alert with weak evidence to an analyst.

Where would you like to start?

  • For CISOs and SOC managers

    See where automation stops, how every decision is audited and how you turn it on step by step.

    Security and data
  • For SOC analysts

    See the evidence behind every closure on one screen, and confirm or reopen it in seconds.

    The analyst view

The problem

Most alerts are not new.

The same maintenance job, the same scanner, the same misconfiguration raise alerts again every day. Closing them one by one fills an analyst's day while the alert that matters waits in the queue.

  • Recurring false positives

    The alert closed yesterday comes back today. The knowledge stays in an analyst's head and never reaches the system.

  • Lost reasoning

    Why an alert was closed is often not written down, so every shift asks the same question again.

  • Alert fatigue

    The more noise there is, the easier it becomes to miss the real signal.

The approach

Two outcomes. No third option.

Every alert gets one question: can we show, with evidence, that it is safe to close?

CLOSE

Close with evidence

An alert shown to be a known issue or a false positive is closed in the SOAR, with its reasoning written as a comment.

ESCALATE

When in doubt, a human decides

If the evidence falls short, or an error or timeout occurs, the alert goes to an analyst. This is the default.

How it works

Pull, evaluate, decide, learn.

SOC Mastermind runs on the machine where your SOAR is installed and pulls alerts from the SOAR API.

  1. Pull

    The alert is taken from the SOAR. The system talks only to the SOAR and connects to nothing else.

  2. Evaluate

    It checks whether there is enough evidence to show the alert is safe to close.

  3. Decide

    With enough evidence, CLOSE; otherwise, ESCALATE. Every decision is recorded with its reasoning.

  4. Learn

    It benefits from your team's decisions, and that process stays under your control.

Trust principles

Safe first, then fast.

If a security product is going to make a mistake, it should make it on the safe side. Every part of the system is designed around that.

  • Doubt goes to a human

    Insufficient evidence, an error or a timeout sends the alert to an analyst. Fail safe, never fail open.

  • No access

    It never touches endpoints, the network or target systems, and never triggers playbooks. In the SOAR it can only close, comment and tag.

  • Observe first

    The system first only observes and closes nothing. You decide when to turn automation on.

  • Auditable and reversible

    Every decision is recorded with its reasoning. Closures can be reopened one by one or in bulk.

  • Kill switch

    One action stops all automatic closing.

  • Your data stays yours

    With a local model, data never leaves the machine. Installation without an internet connection is supported.

Product

The evidence behind every decision, in plain sight.

When an analyst checks a closure, the reasoning and the evidence are on one screen.

Explore the product

An honest start

It starts quiet and speaks up as it earns trust.

It closes nothing on day one. It begins in observe mode: it records the decision it would make and leaves the alert alone.

A product that wrongly closes an alert in a new customer's first week never gets installed again. That is why we start cautiously.

Targets

We state our targets openly.

The figures below are design targets, not measured results. Each installation measures them on your own data in observe mode.

target

≥ 99.7%

Auto-close precision target

The share of alerts the system marks CLOSE that should indeed be closed.

target

40–55%

Coverage target in a mature installation

The share of alerts that can be safely closed without reaching an analyst.

target

≤ 30 s

Decision time target (p95)

From receiving the alert to writing the decision.

target

≤ 30 s

Target time for an analyst to verify a decision

The evidence is on one screen; no searching required.

How we measure

See it on your own alerts, in observe mode.

From day one the system shows what it would do without closing anything. You make the call.