Shipping

Iceslab v1.0 panel - multi-core VPN infrastructure for operators.

Open Iceslab

icecompany.tech/Icecompany

/ Game security|In development — alpha Q4 2026

An anti-cheat
that respects
the player's machine.

Kernel-mode telemetry where it earns its keep, behavioural detection elsewhere. Transparent about what's collected. No always-on drivers, no scope creep, nothing that survives uninstall. The anti-cheat is a guest, not a tenant.

CurrentlyResearch & spec

What we won't do

What we won't do

Three layers

Three layers,
each earning its keep

L1~5%

Kernel-mode telemetry

Watching for what only the kernel can see.

Signed driver, narrow scope: process handle hijacks, unsigned-module injection, DMA-attack indicators. Loaded when the game launches, unloaded at exit. Source is provided to studios under NDA for audit.

Windows · macOS

L2~35%

Userland telemetry

Memory integrity, syscall patterns, input cadence.

In-process detection: hooks on key syscalls, page-protection checks, input timing statistics that distinguish humans from scripts. Runs as part of the game binary, not a separate service.

Windows · macOS · Linux

L3~60%

Server-side behaviour

The match record tells the truth.

Aim curves, reaction-time distributions, decision quality under contested resources — statistical and ML-assisted analysis of telemetry the game already sends. The cheapest layer to ship and the hardest to evade. Most detections happen here.

All — server-side

Timeline

Where we are
Where we're going

Research projects don't ship on calendars. These are the milestones we're working toward.

  1. ● NowMay 2026

    Research & threat-model spec

    Reading public reverse-engineering reports of Vanguard, BattlEye, EAC. Writing a threat model document with explicit non-goals. Talking to two studios about their cheat reports.

  2. → Q3 2026Aug 2026

    L3 prototype on a partner game

    Server-side behavioural analysis layer running on real match telemetry from a friendly studio. No client integration yet — pure analytics pipeline + dashboard.

  3. → Q4 2026Nov 2026

    Alpha — private, NDA-only

    L2 (userland) added on top of L3. SDK provided to 1–3 select studios under NDA. Source is auditable. Detection categories documented in writing before any ban can fire.

  4. → H1 2027Feb 2027

    L1 kernel module + beta

    Kernel layer added only after L2/L3 prove value. Driver code goes through external security review before any signed build ships. Wider studio program opens.

For game studios

Talk to us before
we open alpha.

Q4 2026 alpha takes 1–3 studios. If your game is competitive, online, and you've outgrown what off-the-shelf Icecompanys give you — or if you've watched players uninstall over them — let's have a conversation. Early signal shapes the product.

What an intro call covers

  1. 01Your game's threat model — what cheats you're actually seeing today
  2. 02Which of our three layers fits, and which would be over-engineering
  3. 03Integration cost — honest hours, not a sales estimate
  4. 04Pricing range and what an alpha partnership looks like

What we don't do on intro calls

  1. 01Generic decks — no slideware
  2. 02"Schedule a follow-up" loops — we close out with a written next step
  3. 03Over-promise on detection rates we haven't proven yet
  4. 04Lock you in — NDA goes both ways, no exclusivity tricks

Replies come from a real engineer, not a sales relay. Usually within a working day.