Verisco
VERISCO Alertio · SIEM & SOAR

The layer that decides what deserves a human.

A security operations platform built on Wazuh. It scores every alert against a transparent model, correlates scattered events into kill-chain incidents, and carries response all the way to the endpoint.

verisco · alertio / triage queue
Triage queue scored · gate 50
RECON2
CRED2
EXPLOIT1
PRIV1
LATERAL
IMPACT
TimeAlertHostSeverityScore
09:52Domain admin group modifiedDC-01critical94
09:31SQL injection on web tierWEB-02high74
09:20Kerberoast request burstDC-01high71
09:14IMAP brute forceMAIL-01medium58
08:58Port scan from external hostGW-01low31
Fig. 0.1 — The Alertio console — the triage queue, with the score breakdown for the top alert.
26 Operator screens
148 API endpoints
351 Backend tests
7 Correlation patterns

Counted from the repository. Every screen renders live data — no demo mode, no fixtures.

The problem

Detection is solved. Triage is not.

Wazuh detects well and triages badly. It produces thousands of undifferentiated alerts a day, ranked only by a rule severity that knows nothing about which machine was hit, whether the same thing already fired forty times, or whether this event is one step in a longer attack.

Alertio does not replace the detection engine. It sits on top and adds what a rule severity cannot know.

Transparent scoring
Four weighted factors with the arithmetic shown per alert, and the weights tunable from the interface.
Correlation
Seven patterns, expressed as data rather than code, assembling events into a kill-chain story across six phases.
Response
Eight action types, from a one-click block to signed scripts deployed to the fleet.
One interface
A single time filter, one severity language, and one table-and-drawer pattern across all twenty-six screens.
Correlation

A low score is still surfaced when it belongs to an attack.

Seven patterns assemble scattered events into a kill-chain story. Extending the library is authoring, not engineering.

Threaded by identity, not IP
An attack that crosses network zones from different vantage points would split into two unrelated incidents under an IP-keyed correlator. Alertio joins on the compromised account.
Nothing is dropped
Alerts below the queue gate are suppressed, not deleted. When one turns out to be a step in a longer chain, it is still there to be pulled back in.
Mapped to ATT&CK
Techniques map to kill-chain phases, and a coverage matrix shows fourteen tactics and ninety-seven techniques against what the deployment can actually see.
alertio / att&ck coverage
deep
partial
none
Fig. 1 — Coverage density per tactic, against what the deployment can actually see.
The lab

One chained attack across a real segmented network.

Four isolated demonstrations would show four alerts. One chain shows the product. Alertio was run against a lab with a perimeter, a DMZ, a corporate domain and an out-of-band SOC — and a seven-stage intrusion driven through it end to end.

Internet untrusted · 203.0.113.0/24
Attackerexternal threat actor
DMZ public edge · 10.10.10.0/24
IDS gatewaynetwork intrusion detection
Web serverpublic web application
Mail serverpublic MX and webmail
Corporate LAN crown jewels · 10.20.20.0/24
Domain controllerdirectory and DNS
Employee PCdomain-joined pivot
SOC out-of-band · separate management network
Detection engineevent store and rules
Alertioscoring, correlation, response
Fig. 2 — The demonstration network. Zones light as the selected stage touches them. The SOC is out-of-band — the attacker never sees it.

Stage A — Reconnaissance

Scans the public edge for reachable services.

SensorNetwork IDS
ATT&CKT1595
PhaseRECON

Stage B — Initial access

Exploits the public web application — SQL injection and path traversal.

SensorWeb + network IDS
ATT&CKT1190
PhaseEXPLOIT

Stage C — Delivery & credentials

Phishes a finance user, then brute-forces the mailbox.

SensorMail server
ATT&CKT1566 · T1110
PhaseCREDENTIAL

Stage D — Foothold

Signs in to the domain with the phished finance credentials.

SensorDirectory
ATT&CKT1078
PhaseVALID ACCOUNTS

Stage E — Credential access

Password spray and Kerberoast from the compromised workstation.

SensorDirectory
ATT&CKT1110 · T1558
PhaseCREDENTIAL

Stage F — Privilege escalation

Adds the finance account to Domain Admins.

SensorDirectory
ATT&CKT1484
PhasePRIV_ESC

Stage G — Impact

Mass file modification on a monitored share.

SensorFile integrity monitoring
ATT&CKT1486
PhaseIMPACT

Seven events, four sensors, two zones — one incident.

The attack crosses two zones from different vantage points, so the source address changes partway through. A correlator keyed on IP would break the chain in the middle and report two unrelated incidents. Alertio threads it by the compromised account instead — the same identity is phished on the mail server, used to sign in to the domain, and escalated to Domain Admin.

Response

The decision reaches the endpoint.

Eight action types, from a one-click block through multi-step playbooks to authoring, signing and deploying custom response scripts to the fleet.

Guarded by design
Every action previews before it executes and requires an explicit acknowledgement. Destructive actions — isolate a host, kill a process, quarantine a file — are restricted to administrators.
Human-triggered, deliberately
Playbooks are never fired automatically. The engine supports it; the policy does not. For a platform that can isolate a host, that is a defensible position rather than a missing feature.
Signed and reconciled
Custom scripts are validated, packaged into signed bundles and pushed to agents, with a background job polling the manager for true per-target deployment state rather than assuming success.
Stated plainly

What Alertio does not do.

A product that hides its limits is harder to trust than one that names them.

Notifications are in-app
There is no outbound email, Slack or webhook alerting. Alerts reach analysts inside the interface.
One organisation per deployment
Single tenant by design, on SQLite. The repository layer isolates the change when that stops being the right answer.
The AI is advisory
Triage verdicts come from a security-tuned language model and are recorded as advice an analyst can override — not a classifier, and not the decision.
Wazuh is the implemented source
Integrations sit behind a common interface, but Wazuh is the one that is wired today.

See Alertio against your own environment.

Tell us what you already run for detection and we will tell you what Alertio would add on top of it.