Fighting Game · Guides & deep dives

V2 Anti-Cheat Architecture And Operations

V2 anti-cheat uses layered server authority with client-side evidence as supporting signal.

3sections7 minread

On this page

This runbook covers TODOS.phase-72.72.10.2.15. It is the source-of-truth documentation for V2 anti-cheat architecture, detection methods, and operational procedures across client-side collectors, server-side validators, fair-play review, third-party providers, and CI testing.

Architecture#

V2 anti-cheat uses layered server authority with client-side evidence as supporting signal. Client collectors never own match truth by themselves; they produce signed telemetry, forensic evidence ids, and fair-play signals that the server, replay system, and reviewer tooling can correlate with authoritative match state.

The architecture has these control planes:

  • Client collectors: memory integrity scanning, process scanner, driver integrity, client heartbeat, anti-cheat module obfuscation, screenshot or video capture on suspicion, self-update signatures, hardware fingerprinting, overlay or hook detection, anti-cheat mode opt-in, and minimal-footprint mode.
  • Server validators: server-authoritative movement validation, hit detection, damage validation, economy validation, statistical anomaly detection, and replay-based cheat review.
  • Fair-play adjudication: ban system, appeal workflow, shadow ban routing, player-submitted reports, trusted player fast-track review, analytics dashboard, and ranked fail-closed policy.
  • Provider adapters: EasyAntiCheat and BattlEye integrations are exposed through the provider-neutral third-party anti-cheat integration API. Provider events are normalized before they can mirror into fair-play review.
  • Testing and release gates: the anti-cheat testing framework simulates known cheat types, verifies expected detections, checks deterministic replay evidence, rejects excessive false positives, and emits fair-play validation signals in CI.

The canonical signal flow is:

  1. A detector or provider emits a typed detection event with account, match, build, detector, risk, and evidence identifiers.
  2. Server validators attach authoritative state such as movement replay, lag-compensated hit traces, economy transaction history, or replay review timestamps.
  3. Evidence is normalized into a fair-play signal with a forensic evidence id, privacy classification, and reviewer visibility policy.
  4. Analytics dashboards aggregate detection rate, ban count, appeal rate, and new-cheat emergence metrics without exposing raw player identifiers.
  5. Review and discipline systems apply ban, appeal, shadow-ban, or no-action outcomes according to ranked mode, evidence quality, appeal route, and regional privacy constraints.

Ranked and tournament modes fail closed when required detectors, provider plugins, replay evidence, or fair-play signal routing are unavailable. Non-ranked and single-player modes can use anti-cheat opt-out or minimal-footprint policy when the mode contract allows mods and does not affect competitive integrity.

Detection Methods#

Detection methods are grouped by source, but all methods must be explainable in review evidence and reproducible through replay or test fixtures where possible.

Client-Side Methods#

  • Memory integrity scanning detects injected DLLs, modified code pages, hooked functions, and unexpected executable-page hashes. Findings map to EV2FairPlaySignalType::MemoryIntegrity.
  • Speed hack detection compares signed client game time with server wall-clock time and flags sustained divergence above the configured tolerance.
  • Aimbot heuristics analyze input patterns, snap-to-target speed, perfect tracking, reaction-time windows, target switching, and replay-visible aim correction.
  • Wallhack mitigation combines server-side visibility culling with client-side graphics driver hook detection. The server should not transmit occluded entity state that the client does not need.
  • Movement, teleportation, and fly-hack validation replay client inputs against server physics constraints and rejects impossible position, velocity, or acceleration deltas.
  • Process scanning detects known cheat tools, debuggers, memory editors, and open game-memory handles while preserving an allowlist for signed overlays.
  • Driver integrity verifies graphics driver identity, publisher, signature, and hook status before ranked admission.
  • Client heartbeat requires periodic signed proof that the client build, module hashes, provider session, and heartbeat channel are still intact.
  • Overlay or hook detection identifies render-pipeline injection and separates allowed accessibility, capture, and platform overlays from suspicious hooks.
  • Screenshot or video capture is suspicion-triggered, audit logged, reviewer gated, and subject to privacy redaction before export.
  • Hardware fingerprinting supports repeat-offender correlation and hardware ban enforcement, but must use the documented privacy and appeal guardrails.
  • Self-update signatures allow detection rules to ship without a full game patch. Signature bundles must be signed, versioned, rollback-capable, and tested before release.

Server-Side Methods#

  • Server-authoritative movement validation replays input streams and compares resulting positions to client reports.
  • Server-authoritative hit detection validates shots against server-side player positions with lag compensation and weapon timing.
  • Damage validation rejects impossible damage, fire-rate, range falloff, resource, cooldown, and state transitions.
  • Resource and economy validation treats the server ledger as authority for all currency, item, crafting, and reward transactions.
  • Statistical anomaly detection flags players whose ranked behavior, accuracy, headshot rate, K/D, economy gains, or match graph patterns deviate from population baselines.
  • Replay cheat review records match replays, timestamps suspicious plays, attaches reviewer notes, and keeps evidence available for ban appeal.
  • Third-party anti-cheat provider callbacks map EasyAntiCheat and BattlEye session start, heartbeat, violation, sanction, and session-end events into the V2 fair-play signal model.
  • The testing framework simulates memory tamper, speed hack, aimbot, wallhack, movement hack, process tool, driver hook, and heartbeat tamper fixtures to verify that detection continues to work.

Operational Procedures#

Release Preflight#

Run the focused gates before an anti-cheat release, self-update signature push, or ranked-policy change:

bash
python3 V2/ue/Tools/check-v2-online-anti-cheat-fair-play.py
python3 V2/ue/Tools/check-v2-anti-cheat-report-system.py
python3 V2/ue/Tools/check-v2-trusted-player-program.py
python3 V2/ue/Tools/check-v2-anti-cheat-analytics-dashboard.py
python3 V2/ue/Tools/check-v2-third-party-anti-cheat-integration.py
python3 V2/ue/Tools/check-v2-anti-cheat-testing-framework.py
python3 V2/ue/Tools/check-v2-process-scanner.py
python3 V2/ue/Tools/check-v2-driver-integrity.py
python3 V2/ue/Tools/check-v2-client-heartbeat.py
python3 V2/ue/Tools/check-v2-anti-cheat-module-obfuscation.py
python3 V2/ue/Tools/check-v2-suspicion-capture.py
python3 V2/ue/Tools/check-v2-anti-cheat-self-update.py

Release approval requires signed detector configuration, current provider SDK metadata, replay evidence coverage, fair-play review routing, appeal route coverage, and analytics dashboard visibility. Ranked releases must verify fail-closed behavior before enabling matchmaking.

Signature And Provider Rollout#

Detection signatures roll out in stages:

  1. Generate and sign the self-update bundle.
  2. Run the anti-cheat testing framework against known cheat fixtures and benign false-positive fixtures.
  3. Canary non-ranked telemetry first when the detector can operate in observation mode.
  4. Enable ranked enforcement only after detection recall, false-positive rate, appeal-rate baseline, and replay-evidence coverage are within policy.
  5. Keep rollback bundles available until the new-cheat emergence dashboard is stable for the release window.

Third-party provider rollout follows the provider-neutral API. EasyAntiCheat and BattlEye plugin readiness, signed session tickets, heartbeat channel status, server callback signatures, and fallback fail-closed policy must pass before ranked admission.

New Cheat Emergence#

When analytics detects a new cheat emergence signal:

  1. Triage detection rate, ban count, appeal rate, affected mode, platform, region, and provider session distribution.
  2. Preserve representative replay evidence, telemetry snapshots, and detector evidence ids.
  3. If evidence is weak, route to manual fair-play review instead of automatic discipline.
  4. Build a known-cheat fixture for the anti-cheat testing framework.
  5. Update signatures or provider mappings, rerun release preflight, and canary before ranked enforcement.
  6. Monitor appeals and false-positive reports after rollout.

Review, Discipline, And Appeals#

Player-facing discipline must be traceable to evidence. A reviewer should be able to see match id, timestamp, detector id, provider event id when present, replay segment, risk score, account history, and prior appeal outcomes.

Ban waves are allowed when evidence quality is consistent and reviewer-approved. Shadow-ban routing is reserved for confirmed cheater pool isolation and must not hide evidence from the appeal workflow. Hardware bans require stronger evidence, privacy review, and explicit appeal coverage.

Appeals must preserve the original evidence bundle, reviewer notes, detector version, provider session id, and final reviewer decision. If a detector or signature is rolled back for false positives, affected sanctions are requeued for review.

Evidence, Privacy, And Access#

Anti-cheat evidence uses least-privilege access:

  • Raw screenshots, video, hardware fingerprints, and provider payloads are restricted to fair-play operations and privacy-approved tooling.
  • Analytics dashboards use redacted player identifiers and aggregate slices by mode, platform, region, detector, and provider.
  • Evidence exports include audit id, reason, reviewer id, timestamp, and retention classification.
  • Minimal-footprint and opt-out modes must never feed ranked enforcement.
  • Classifier and statistical outputs are review-prioritization signals unless a separate policy explicitly authorizes automated action.

Incident Response#

For active bypass incidents, open an anti-cheat incident ticket with detector, mode, build, region, provider, replay, and appeal context. Disable ranked admission for affected detector families if required dependencies are missing. Preserve logs and replay artifacts before changing signatures. Coordinate with privacy, platform, security, and fair-play review owners before expanding collection or discipline scope.

Verification Inventory#

The documentation contract is V2/ue/Content/V2/Online/AntiCheatDocumentation_V2_Contract.json, and the focused checker is V2/ue/Tools/check-v2-anti-cheat-documentation.py.

The documentation is anchored to these implementation contracts:

  • V2/ue/Content/V2/Online/OnlineAntiCheatFairPlay_V2_Contract.json
  • V2/ue/Content/V2/Online/AntiCheatReportSystem_V2_Contract.json
  • V2/ue/Content/V2/Online/TrustedPlayerProgram_V2_Contract.json
  • V2/ue/Content/V2/Online/AntiCheatAnalyticsDashboard_V2_Contract.json
  • V2/ue/Content/V2/Online/ThirdPartyAntiCheatIntegration_V2_Contract.json
  • V2/ue/Content/V2/Online/AntiCheatTestingFramework_V2_Contract.json
  • V2/ue/Content/V2/Online/ProcessScanner_V2_Contract.json
  • V2/ue/Content/V2/Online/DriverIntegrity_V2_Contract.json
  • V2/ue/Content/V2/Online/ClientHeartbeat_V2_Contract.json
  • V2/ue/Content/V2/Online/AntiCheatModuleObfuscation_V2_Contract.json
  • V2/ue/Content/V2/Online/SuspicionCapture_V2_Contract.json
  • V2/ue/Content/V2/Online/AntiCheatSelfUpdate_V2_Contract.json

Required release gates:

  • anti-cheat-documentation-source-present
  • anti-cheat-documentation-architecture-coverage
  • anti-cheat-documentation-detection-methods
  • anti-cheat-documentation-operational-procedures
  • anti-cheat-documentation-evidence-privacy
  • anti-cheat-documentation-ci-wired