WorldPreventer – 高级反作弊与世界保护
⚠️ 早期开发通知
WorldPreventer 仍处于非常早期的开发阶段。许多系统已具备功能,但机器学习 (ML) 检测系统不附带经过训练的模型。
- 不要启用 ML 检测,除非你已训练了自己的模型。
- 如果你看到通知
REQUEST_STATS observed (WorldTools signature),这很可能表明某个客户端正试图下载你的世界。
概述
WorldPreventer 是一款高性能的反作弊与世界保护插件,旨在保护你的服务器免受各种威胁,包括:
它结合了数据包级监控、行为分析和自适应响应,以保护你的服务器免受恶意活动侵害。
特性
核心保护系统
-
区块利用漏洞防护
- 检测并防止快速加载/卸载区块
- 监控区块交互模式
- 阻止区块边界利用漏洞
-
WorldTools 拦截器
- 阻止 WorldTools 客户端功能
- 防止未经授权的世界下载
- 检测并阻止快速加入/离开模式
-
深度数据包检测 (DPI)
- 分析网络流量中的可疑模式
- 检测并阻止恶意数据包序列
- 监控数据量与时序异常
-
自适应响应系统
- 动态威胁等级评分
- 逐步升级的响应措施(警告 → 踢出 → 封禁)
- 完全可配置的操作
机器学习检测(可选)
重要提示: ML 系统默认处于禁用状态,且不附带经过训练的模型。请勿在自行训练模型前启用它。
正确配置后,ML 系统将提供:
- 异常检测 – 发现不寻常的玩家行为
- 行为分析 – 学习正常的服务器模式
- 实时评分 – 持续评估玩家操作
- 自适应阈值 – 根据服务器状况进行调整
追踪特征(共 12 项):
区块转换、容器交互、移动体积、实体生成速率、数据包爆发、系统性移动、加入激增、客户端品牌、时间精度、数据量方差、交互模式、移动熵。
命令与权限
管理员命令
/worldpreventer reload – 重新加载配置
/worldpreventer status – 查看保护状态
/worldpreventer simulate <player> <test> – 运行模拟测试
/worldpreventer ml [enable|disable|status] – 管理 ML 系统
权限节点
worldpreventer.admin – 完全访问权限及绕过检测
worldpreventer.staff – 接收可疑活动警报
worldpreventer.bypass – 将玩家排除在检测之外
配置
示例 config.yml:
# 加入后的宽限期
join_grace_period_seconds: 120
# 升级操作
actions:
on_exceed: WARN # WARN, KICK, BAN
# 嫌疑评分
scoring:
enabled: true
decay_per_minute: 10.0
thresholds:
warn: 100.0
kick: 500.0
# 机器学习(默认禁用)
ml:
enabled: false
direct_kicks: false
score_threshold: 3.0
points_on_exceed: 20.0
kick_message: "§cKicked by ML anti-cheat system"
window_seconds: 60
autosave_minutes: 5
decay: 0.98
# 基于 ProtocolLib 的数据包拦截
packet_blocking:
enabled: true
block_stats_request: false
block_statistics: false
block_map_data: false
corrupt_block_entities: false
inject_fake_data: false
集成
依赖项
- 必需: ProtocolLib
- 推荐: Vault
API 示例
WorldPreventer plugin = (WorldPreventer) Bukkit.getPluginManager().getPlugin("WorldPreventer");
// 获取嫌疑分数
double score = plugin.getSuspicionScorer().getScore(player.getUniqueId());
// 添加自定义检测点数
plugin.getSuspicionScorer().addScore(player.getUniqueId(), 10.0, "CUSTOM_EVENT");
工作人员通知
拥有相应权限的工作人员会收到以下警报:
- 可疑活动
- ML 系统检测结果
- 配置更改
- 警告与错误
绕过选项
以下玩家将被排除在检测之外:
- 持有
worldpreventer.bypass 权限的玩家
- 处于创造模式的玩家
- 处于旁观者模式的玩家
支持
如需报告 Bug、提交功能请求或寻求技术支持,请访问 GitHub 仓库。
WorldPreventer – Advanced Anti-Exploit & World Protection
⚠️ Early Development Notice
WorldPreventer is still in a very early stage of development. Many systems are functional, but the machine learning (ML) detection system does not ship with a trained model.
- Do not enable ML detection unless you have trained your own model.
- If you see the notification
REQUEST_STATS observed (WorldTools signature), this likely indicates that a client is attempting to download your world.
Overview
WorldPreventer is a high-performance anti-exploit and world protection plugin designed to safeguard your server from a wide range of threats, including:
- World downloads
- Chunk exploitation
- Suspicious client behavior
It combines packet-level monitoring, behavioral analysis, and adaptive responses to protect your server from malicious activity.
Features
Core Protection Systems
-
Chunk Exploit Prevention
- Detects & prevents rapid chunk loading/unloading
- Monitors chunk interaction patterns
- Stops chunk boundary exploits
-
WorldTools Blocker
- Blocks WorldTools client functionality
- Prevents unauthorized world downloads
- Detects & blocks rapid join/leave patterns
-
Deep Packet Inspection
- Analyzes network traffic for suspicious patterns
- Detects & blocks malicious packet sequences
- Monitors data volume & timing anomalies
-
Adaptive Response System
- Dynamic threat-level scoring
- Escalating responses (warn → kick → ban)
- Fully configurable actions
Machine Learning Detection (Optional)
Important: The ML system is disabled by default and ships without a trained model. Do not enable it unless you train your own.
When configured properly, the ML system provides:
- Anomaly Detection – Spotting unusual player behavior
- Behavioral Analysis – Learning normal server patterns
- Real-time Scoring – Constantly evaluating actions
- Adaptive Thresholds – Adjusting to server conditions
Tracked Features (12 total):
Chunk transitions, container interactions, movement volume, entity spawn rates, packet bursts, systematic movement, join surges, client brand, timing precision, data volume variance, interaction patterns, movement entropy.
Commands & Permissions
Admin Commands
/worldpreventer reload – Reload configuration
/worldpreventer status – View protection status
/worldpreventer simulate <player> <test> – Run simulation tests
/worldpreventer ml [enable|disable|status] – Manage ML system
Permissions
worldpreventer.admin – Full access & bypass
worldpreventer.staff – Receive suspicious activity alerts
worldpreventer.bypass – Exclude player from detection
Configuration
Example config.yml:
# Grace period after join
join_grace_period_seconds: 120
# Escalation actions
actions:
on_exceed: WARN # WARN, KICK, BAN
# Suspicion scoring
scoring:
enabled: true
decay_per_minute: 10.0
thresholds:
warn: 100.0
kick: 500.0
# Machine Learning (disabled by default)
ml:
enabled: false
direct_kicks: false
score_threshold: 3.0
points_on_exceed: 20.0
kick_message: "§cKicked by ML anti-cheat system"
window_seconds: 60
autosave_minutes: 5
decay: 0.98
# ProtocolLib-based packet blocking
packet_blocking:
enabled: true
block_stats_request: false
block_statistics: false
block_map_data: false
corrupt_block_entities: false
inject_fake_data: false
Integration
Dependencies
- Required: ProtocolLib
- Recommended: Vault
API Example
WorldPreventer plugin = (WorldPreventer) Bukkit.getPluginManager().getPlugin("WorldPreventer");
// Get suspicion score
double score = plugin.getSuspicionScorer().getScore(player.getUniqueId());
// Add custom detection points
plugin.getSuspicionScorer().addScore(player.getUniqueId(), 10.0, "CUSTOM_EVENT");
Staff Notifications
Staff with permission receive alerts for:
- Suspicious activity
- ML system detections
- Config changes
- Warnings & errors
Bypass Options
Players excluded from detection:
worldpreventer.bypass permission holders
- Players in Creative mode
- Players in Spectator mode
Support
For bug reports, feature requests, or support, please visit the GitHub repository.
评论 0
登录后参与评分
还没有评论,来抢沙发吧