Run PROTEKT on your campus

Everything your institution's network team provides, in one pass.

Two things go onto your estate: a lightweight PROTEKT Agent on the endpoints you choose, and a passive PROTEKT Network Sensor at each internet edge. Everything else — detection, correlation, triage, response — runs in the PROTEKT SOC. Tell us who you are and this page becomes your provisioning brief.

Nothing is submitted from this page. Your entries personalise the brief — and set the sensor count and sizing tier — you send us at the end.

Transmits nothing onto the network

The sensor's capture interface has no IP address and never sends a packet. Agents speak outbound-only, over encrypted channels, to the PROTEKT SOC.

Cannot block or modify traffic

The sensor inspects a mirrored copy — it has no position in the traffic path. Response actions are decisions made with your team, never automatic network intervention.

Fails safe

If the sensor, an agent, or an uplink fails, the only consequence is a gap in visibility. Nothing PROTEKT deploys can cause a network or endpoint outage.

Plane one — Endpoint telemetry

A · The PROTEKT Agent

Your team deploys · we operate

A lightweight service installed on the endpoints your institution chooses to protect — labs, staff machines, servers. It watches the machine, not the person: telemetry goes outbound over an encrypted channel to the PROTEKT SOC, and the footprint is small enough to be invisible on shared lab hardware.

ItemDetail
Supported systemsWindows 10/11 and Windows Server 2016+ · macOS · major Linux distributionsMixed estates are normal — most campuses are Windows-majority with Linux servers and macOS pockets
FootprintBackground service — typically under 1% CPU and ~100 MB memoryNo reboot required to install. No user-visible interface on the endpoint.
DeploymentYour standard software distribution: Group Policy, Intune, SCCM/MECM, Jamf, or scripted install for Linux fleetsEnrollment credentials and packages are issued at onboarding — one package per device group
ConnectivityOutbound only, encrypted, to the PROTEKT SOCExact egress details ship with the enrollment package; nothing inbound to any endpoint

What the agent observes

  • System and security event logs
  • Critical file integrity changes
  • Process and service activity
  • Configuration and patch posture
  • Known-vulnerable software present on the device

What it never does

  • No keystroke capture
  • No screen recording
  • No reading or uploading of personal file contents
  • No browsing surveillance
  • No access for anyone outside the PROTEKT SOC and your named contacts
Plane two — Network telemetry

B · The sensor host

Your team provides

One virtual machine per internet breakout, on your existing virtualization platform. Sizing is driven by mirrored traffic volume, not device count — the table below adjusts to the edge capacity you selected at the top of the page, so what you see is your tier.

ResourceSpecification
Sizing tierStandardDefault — refine by selecting your internet edge at the top of the page
vCPU8Multi-threaded inspection headroom for the tier's traffic band
Memory32 GBFlow tables and inspection buffers scale with traffic volume
Disk200 GBThin-provisioned; alert-only output is small — standard shared storage (NAS/SAN) is fine
Operating systemUbuntu Server 24.04 LTSWe install onto an empty VM with console access, or your team deploys from a standard template
Interface 1 — managementRouted server network, static IPOutbound-only; exact egress rules in section D
Interface 2 — captureConnected to the mirror destination (section C)No IP address assigned. Passive receive only.
Plane two — Network telemetry

C · The traffic mirror

Your team provides

One SPAN / port-mirror session copying your internet-edge traffic to the sensor. The single most important choice: mirror the internal (LAN-side) interface of the perimeter firewall — before address translation — so every alert names a real internal host instead of a translated address. That is the difference between an actionable alert and a useless one.

ParameterRequirement
SourceLAN-side interface of the perimeter firewall — the campus↔internet path, pre-translation
DirectionBoth transmit and receive on the source portA one-directional mirror blinds half of every conversation
DestinationThe port patched to the sensor's capture interface
VLAN tagsPreserve (802.1Q) where the platform supports itStripped tags are workable; preserved tags give per-segment attribution
Scope, phase oneInternet-edge path onlyInternal and inter-VLAN links are a phase-two conversation — one clean mirror beats three noisy ones
Do I need more than one sensor?One sensor per traffic chokepoint — not per subnet or VLAN. The edge mirror above already carries every segment's traffic to and from the internet, so one sensor covers the whole campus's north-south path no matter how many VLANs it's carved into. Movement between internal segments is watched from the endpoint side by the PROTEKT Agents, and network-level internal mirrors are a phase-two addition to the same sensor, not new ones. Additional sensors are needed only for additional internet breakouts or separate campuses — which is why we ask for that count at the top of this page.
Not sure your network can mirror?Most modern campus switches and firewalls support port mirroring (SPAN). If yours doesn't — or you're not certain — don't stop here: alternative deployment patterns exist. Email hello@protektor.cloud before provisioning and we'll design around it.

D · Network access

Your team provides

Everything PROTEKT deploys speaks outbound only. Nothing inbound from the internet is required — the sensor and every agent originate their own encrypted connections to the PROTEKT SOC.

FromProtocol / PortDestinationPurpose
Sensor VMUDP 51820 outboundPROTEKT SOC endpoint address provided at onboardingEncrypted telemetry tunnel
Sensor VMTCP 443 / 80 outboundOS and detection-content repositoriesSystem and detection rule updates
Sensor VMUDP 123 outboundYour NTP source, or a public poolTime synchronization — accurate timestamps are what make correlation and investigation possible. If your NTP source is internal, ensure this VM is permitted to reach it.
Agent networksEncrypted, outboundPROTEKT SOC details issued with the enrollment packageEndpoint telemetry

E · Six answers

Your team answers

Answer these right here — they flow straight into your brief below. Each question says why we ask and where the answer usually lives, and every card has a note field if anything needs explaining. Question six is the one that most often moves a go-live date.

01Internet link capacity and typical peak utilizationOPEN
Confirms the sensor sizing — if sustained traffic exceeds 10 Gbps, we adjust before install, not after.
Where to find it: your ISP agreement, or the throughput graph on your edge firewall / router.
Edge capacity
Typical peak utilization
Add a note
02Make and model of the device providing the mirrorOPEN
Determines mirror capabilities: VLAN preservation, session limits, remote-mirroring support.
Where to find it: your network diagram or asset register — the switch or firewall the mirror will come from.
Add a note
03Is a virtualization host with a free physical NIC within patching distance of that device?OPEN
The preferred design patches the mirror straight into a dedicated NIC on the host.
Who to ask: your virtualization / data-centre team — a host in the same rack or row with an unused NIC.
Add a note
04Does the network support remote mirroring (RSPAN / ERSPAN) to where a host lives?OPEN
With no adjacent host, remote mirroring carries the copy across the network instead — workable, with a slightly different configuration on both sides.
Who to ask: your network team — RSPAN/ERSPAN capability on the path between the mirror device and the host.
Add a note
05Primary and secondary operational contactsOPEN
Who receives alert notifications, and who we call when the sensor, an agent fleet, or a telemetry feed needs attention.
Add a note
06Does the mirror configuration require change-window approval?OPEN
Approval lead time, not technical work, is what usually threatens the date — if a change-advisory cycle is involved, lodge the request the week provisioning begins.
Who to ask: your change management process — does a SPAN / mirror change need approval, and what is the usual lead time.
Add a note

F · What happens, and when

Typical: five working days

From provisioning to live telemetry. Once the host, mirror, and first device groups exist, everything inside them is ours for the duration of the engagement: install, tuning, detection-content updates, software lifecycle, and 24/7 operation from the PROTEKT SOC.

Day 0
Brief received; change request lodged if required
Your team
Day 1–3
VM + mirror provisioned; first agent groups staged
Your team
Day 3–4
Sensor installed, tunnel live, agents enrolling, tuning pass
Protekt
Day 5
Joint validation and sign-off
Both

Validation is defined, not vague: a benign test detection is triggered on the network path and on an enrolled endpoint, both alerts are confirmed visible in your institution's PROTEKT portal with correct host attribution and accurate timestamps, and a written sign-off note records it. That note is the line between "provisioned" and "protecting."

For validation day, have ready: one enrolled endpoint to trigger the benign test, access to your PROTEKT portal during the session, and your primary contact available for a 30-minute call.

The last step

Your deployment brief

Generated from this page

This is what the deployment looks like from your institution's side — built from your entries and your checklist. Send it to us before provisioning and we come back with the enrollment package, the SOC endpoint details, and a named engineer.

Deployment brief — your institution
Endpoints in scope
Internet edge
Network sensors
1
Sensor tier
Standard
Provisioning items
0 of 0 complete
Primary contact
— to be confirmed
Derived: select your internet edge and breakout count at the top of the page.

Your six answers

    Outstanding before install

      Copied — paste anywhere