Security • Competitive Integrity

Play fair with Ember Anti-Cheat

Real-time, server-side protection for every competitive match on C². Ember Anti-Cheat analyzes signals continuously, flags anomalies, and preserves the ladder’s integrity — without getting in your way.

Live integrity checks Server-side detection Privacy-first design Transparent audit trails
Ember Anti-Cheat

How it works

Queue & Match

During matchmaking, Ember runs passive checks — client version, process fingerprint, and low-noise device traits. Nothing intrusive; everything is hashed and ephemeral.

Account identity stitching

Risk-aware linking for email, provider, and profile history keeps smurfs in check.

Light reputation cache

Soft risk from recent bans/timeouts travels with the account for tempered matching.

Runtime Signals

Live telemetry with strict budgets; everything is aggregated and anonymized.

Reaction cadence

Natural latency bands vs. robotic bursts—time-series catches macros & scripts.

Windowed stability

Short windows detect jitter anomalies while long windows smooth out noise.

Inference Model

Multi-feature scoring with explainability—no single metric ever bans.

Robust z-scores

Distribution-aware thresholds make score spikes stand out without catching grinders.

Integrity Actions

Graduated responses keep queues healthy without false-positive whiplash.

Soft flags

Queue-neutral markers adjust match confidence and route to human review if needed.

What raises risk?

Toggle suspicious patterns to see how evidence accumulates. Multiple weak signals → strong confidence.

0%
Low risk

Ember correlates many more signals than shown here. This is a simplified visualization for education.

Integrity checks

Background validation during queue & match to prevent foul play without interrupting legitimate sessions.

Detection & actions

Automated flags with manual escalation for edge cases. Penalties scale with confidence and recency.

Audit trails

Transparent timelines for appeals: what we saw, when we saw it, and why an action occurred.

Privacy & fairness

Built to be effective and respectful.

Only interaction shapes and metadata: timing, focus state changes, version hashes, device & session consistency. We do not store question content, audio, or your screen.

No. Models adapt to ladder pace and your personal baseline. Single spikes won’t penalize; sustained impossible patterns across multiple signals will.

Every action writes to an audit trail. Appeals are reviewed by humans with the same evidence the system used. Reversals restore points/standing when applicable.