Skip to content

Features

Jump to bottom
Wiki Generator edited this page Mar 18, 2026 · 1 revision

Features

OverWatch-ML provides a comprehensive suite of cheat detection features powered by machine learning.

Core Detection Features

X-Ray Detection

The primary feature - detecting players using x-ray or similar cheats to find hidden ores.

How it works:

  • Monitors block break events for valuable ores
  • Tracks ore-to-stone ratios for each player
  • Analyzes mining patterns and trajectories
  • Uses decoy ore system to catch confirmed x-ray users
  • Employs ML models to detect anomalous behavior

Detected Ores:

  • Diamond Ore
  • Deepslate Diamond Ore
  • Gold Ore
  • Deepslate Gold Ore
  • Emerald Ore
  • Deepslate Emerald Ore
  • Ancient Debris

Combat Cheat Detection

Advanced heuristic analysis for combat modifications.

Detects:

  • Kill Aura / Aimbot
  • Reach extensions
  • Invalid angle changes
  • Unusual attack patterns
  • Speed modifications

Machine Learning System

Three-Tier ML Pipeline

OverWatch-ML uses a sophisticated three-tier detection system:

Tier 1: Real-Time Heuristics

  • Processes every player action in real-time
  • Negligible performance impact (<1ms)
  • Filters obvious legitimate behavior
  • Flags suspicious activities for Tier 2

Tier 2: Behavioral Analysis

  • Builds short-term behavioral profiles
  • Analyzes action sequences and timings
  • Identifies patterns common to cheat clients
  • Runs asynchronously (5-20ms)

Tier 3: Deep Machine Learning

  • Uses Isolation Forest algorithms
  • Uses Autoencoder neural networks
  • Compares against trained datasets
  • Runs once every 5 minutes per player

Bot Training System

Train the ML system using AI bots with different behaviors:

Bot Types:

  • NORMAL_MINER - Standard legitimate mining
  • XRAY_CHEATER - Simulates x-ray usage
  • TUNNEL_MINER - Focused tunnel mining
  • RANDOM_MINER - Unpredictable mining
  • EFFICIENT_MINER - Highly efficient mining
  • SURFACE_MINER - Surface-level mining only

Server-Adaptive Learning

The system automatically:

  • Adjusts detection thresholds based on your server
  • Learns from player behavior patterns
  • Updates models with new data
  • Adapts to your specific economy

Additional Features

Decoy Ore System

  • Places fake ores that only x-ray users will find
  • Configurable thresholds and timing
  • Automatic revert after time limit
  • Buried placement for realism

Punishment System

6 levels of escalating punishments:

  1. Warning
  2. Mining debuff
  3. Inventory lock
  4. Area restriction
  5. Temporary ban
  6. Permanent ban

Staff GUI

Comprehensive graphical interface:

  • Player profiles with detailed stats
  • Suspicious player list
  • Real-time alerts
  • ML analysis results
  • Configuration management
  • Punishment controls

Discord Integration

  • Real-time notifications
  • Configurable alert levels
  • Webhook support for custom integrations

Command Hiding

Hide plugin commands from regular players:

  • Shows "Unknown command" instead of "No permission"
  • Useful for security on large servers

Data Management

  • Automatic saving of stats and suspicious players
  • YAML or SQLite storage options
  • Configurable auto-save intervals
  • Export capabilities

Performance

Optimized for minimal server impact:

Metric Value
TPS Impact < 0.5%
Memory Usage 50-200 MB
Tier 1 Processing < 1ms
Tier 2 Processing 5-20ms
Tier 3 Processing Every 5 minutes

Supported Server Software

  • Paper (Recommended) - Full feature support
  • Folia - Full support with multi-threading
  • Spigot - Basic support

Clone this wiki locally