How it works
From first connector to contained threat
Connecting sources takes days. Baselines settle in weeks. After that, Lairhunt hunts continuously and hands your team explained incidents with a recommended next step.
Token theft and mailbox rule creation
- Techniques
- T1078 · T1021.002 · T1003
- Alerts folded in
- 41 → 1 incident
- Entities
- 3 hosts · 2 identities
- Severity / confidence
- High / 0.92
- Login by
svc_backupfrom a subnet it has never used - SMB admin-share access,
fin-db-07tohr-fs-02 - Credential-store memory read on
hr-fs-02 - Veridex links all three into one incident
The process
Six steps from kickoff to steady state
Days 1–5
Connect sources
Add network taps or flow exports, your identity provider, cloud audit logs and EDR telemetry.
Weeks 1–3
Learn baselines
Huntra builds a model of normal per entity, peer group and time of day for your tenant.
Weeks 2–4
Tune with your team
We review early incidents together and mark known-good behaviour so the models settle.
Ongoing
Hunt continuously
Every event is scored. Anomalies that matter become Veridex investigations.
Ongoing
Respond with guardrails
Quellix recommends containment. You decide which actions may run automatically.
Ongoing
Improve from verdicts
Analyst feedback and Synthra simulations keep detection current.
Who does what
Clear ownership from day one
| Task | Your team | Lairhunt |
|---|---|---|
| Telemetry access | Grant read access or point exports at Lairhunt | Provide connectors, collectors and sizing |
| Asset and identity context | Share CMDB tags and crown-jewel lists | Import and keep them in sync |
| Baseline tuning | Confirm what normal looks like | Run the models and surface outliers |
| Incident review | Confirm or dismiss incidents | Investigate, explain and learn from verdicts |
| Guardrails | Set autonomy per action and asset | Enforce them and log every decision |
| Response integrations | Approve API scopes in EDR and IdP | Build, test and monitor the actions |
Guardrails
You decide how much runs on its own
Every containment action has a mode: recommend, approve or auto. Guardrails are a versioned file your SOC lead owns, so a change is reviewed and tracked like code.
- Severity and confidence thresholds per action
- Asset tags that are never touched automatically
- Named approvers for sensitive actions
- Time limits on blocks and isolations
# guardrails.yaml: what Quellix may do without asking version: 3 default_mode: recommend # recommend | approve | auto rules: - action: isolate_host mode: auto when: severity: ">= high" confidence: ">= 0.9" asset_tags_exclude: [domain-controller, prod-db] - action: disable_identity mode: approve approvers: [soc-lead, iam-oncall] - action: block_flow mode: auto max_duration: 4h always: log: true rollback: one_click blast_radius_check: required
What you get back
An incident you can act on
Threat
An explained incident
What happened, in order, with the evidence for each step and the ATT&CK techniques involved.
Action
A recommended response
The smallest action that contains the threat, with its predicted blast radius.
Review
Borderline results flagged
When confidence is low, Veridex says so and asks for a person instead of guessing.
Questions
About onboarding
What telemetry do we need to start?
Identity sign-in logs plus either network flows or EDR events are enough for a first deployment. More sources give Veridex more to pivot on.
How long before incidents are useful?
Baselines usually settle within two to three weeks. Lairhunt marks early incidents as low confidence while it learns.
Does Lairhunt need agents on endpoints?
No. It reads telemetry from the EDR and collectors you already run. Sentix appliances are only for OT and edge networks.
Can we run hunts by hand too?
Yes. Analysts can query full telemetry from the console or API at any time.
Start with one environment
The design partner program covers one environment, your core telemetry sources and eight weeks of hunting.