Vulnerability Disclosure: Critical RCE in MessagePack

Overview

This repository contains a proof-of-concept (PoC) for an Arbitrary Code Execution (ACE) vulnerability identified in the MessagePack deserialization process.

Target Format

.msgpack

Description

The application utilizes msgpack.unpack() on untrusted model/data files without strict type validation. By crafting a payload using MessagePack extension types, an attacker can instantiate and execute arbitrary system commands during the deserialization process.

Proof of Concept

The provided poc_messagepack.msgpack triggers os.system('echo ACE_TRIGGERED').

Steps to Reproduce

  1. Install a vulnerable version of the target parser/library.
  2. Execute the loader:
    import msgpack
    with open("poc_messagepack.msgpack", "rb") as f:
        msgpack.unpack(f)
    
  3. Observe the execution of the command.

Recommendation

Implement a strict allow-list for deserialized objects or transition to non-executable serialization formats.

Security Impact

Full remote code execution on the host machine during the model loading phase.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support