Skip to content

Security: EndlessPixel-Studio/FakePlayer-CE

Security

docs/SECURITY.md

Security Policy

English | 简体中文 | 繁體中文 | 日本語

FakePlayer CE spawns real server-side players and ships an optional HTTP admin API, so a misconfiguration can hand an attacker control over your server. This document covers what is supported, how to report a problem, and how to run the plugin safely.

Supported versions

Only the newest published build receives security fixes. Older builds are not patched — update to the latest fp.buildN from GitHub Releases or Hangar.

Build Supported
Latest published fp.buildN ✅
Any older build ❌

Reporting a vulnerability

Do not open a public issue for a security problem.

Use GitHub's private reporting: open the Security tab → Report a vulnerability. That creates a private advisory only the maintainers can see.

Please include:

  • Plugin build (/version FakePlayer-CE or the jar file name)
  • Server software and version (Paper / Purpur / Leaf / Folia, Minecraft version)
  • Steps to reproduce, expected vs. actual behaviour
  • Relevant config (redact your token) and log excerpts

You will get an acknowledgement as soon as possible; this is a volunteer project, so please allow a few days. Once a fix ships, we are happy to credit you in the advisory and the changelog.

HTTP Admin API

The HTTP API is disabled by default and every endpoint has an independent switch. When you enable it, keep the following in mind.

What the plugin already does for you

Protection Behaviour
Token authentication Authorization: Bearer <token> only. ?token= is deliberately not supported, because query strings leak into proxy logs and browser history. An empty token generates a random one at startup and prints it to the console.
Auth before routing Authentication is evaluated before the per-endpoint switch, so 401 / 403 responses cannot be used to discover which endpoints are enabled.
Constant-time comparison The token is compared with MessageDigest.isEqual, not String.equals.
Log hygiene Tokens are redacted and control characters replaced before anything is logged.
Rate limiting Per source IP, with a lockout after repeated authentication failures.
Host / Origin allowlist allowed-hosts rejects requests whose Host or cross-origin Origin does not match, which mitigates DNS rebinding.
No caching Responses carry Cache-Control: no-store.

What is still on you

  • There is no TLS. The API speaks plain HTTP. Bind it to 127.0.0.1, or put it behind a reverse proxy that terminates TLS, and restrict access at the firewall.
  • /cmd runs commands as a fake player, with that fake player's permissions. It is the most dangerous endpoint — leave http-admin.interface.cmd disabled unless you need it, and never OP a fake player: anyone who can drive that fake player inherits those permissions.
  • Keep the token secret. Treat it like a root password. Rotate it if it ever appears in a screenshot, a log paste, or a public repository.
  • Expose only what you need. Every endpoint has a switch under http-admin.interface; turn off the ones you do not use.

Hardening checklist

http-admin:
  enabled: true
  host: 127.0.0.1          # loopback, or an internal address behind a reverse proxy
  token: ""                # empty generates a random token; set a long random value instead
  allowed-hosts: []        # set your public hostname when exposing through a proxy
  rate-limit:
    requests-per-minute: 120
    auth-failures: 10
    lockout-seconds: 60
  interface:
    cmd: false             # disable anything you do not actually call

If you expose the API beyond localhost, also:

  • terminate TLS at the proxy and forward only from the proxy's address
  • add authentication at the proxy layer (mTLS, basic auth, or an SSO gateway) on top of the token
  • monitor the plugin log for repeated 401 / 429 responses

Scope

In scope: the plugin jar (command handling, HTTP API, NMS modules under versions/, login-plugin compatibility, update checker, Hangar/GitHub publishing workflows in this repository).

Out of scope: issues that require an already-compromised server, deliberately granting a fake player OP rights, running with forced-execution: true on an unsupported server version, and griefing performed through features working exactly as documented (for example /fp cmd).

There aren't any published security advisories