CVE-2024-5480

RCE vulnerability in PyTorch RPC allows arbitrary command execution.

• CVSS 10.0
• CWE-77
• Input Validation and Sanitization
• Remote

A vulnerability in the PyTorch's torch.distributed.rpc framework, specifically in versions prior to 2.2.2, allows for remote code execution (RCE). The framework, which is used in distributed training scenarios, does not properly verify the functions being called during RPC (Remote Procedure Call) operations. This oversight permits attackers to execute arbitrary commands by leveraging built-in Python functions such as eval during multi-cpu RPC communication. The vulnerability arises from the lack of restriction on function calls when a worker node serializes and sends a PythonUDF (User Defined Function) to the master node, which then deserializes and executes the function without validation. This flaw can be exploited to compromise master nodes initiating distributed training, potentially leading to the theft of sensitive AI-related data.

CVE

CVE-2024-5480

CWE

CWE-77

CVSS base score

10.0

Published

2024-06-06

OWASP

A01 Broken Access Control

Orthogonal defect classification

Function

Code defect classification

Incorrect Functionality

Category

Input Validation and Sanitization

Subcategory

Command Injection

Accessibility scope

Remote

Impact

Arbitrary Code Execution

Affected component

PyTorch

Fixed by upgrading

Yes

Solution

Upgrade to PyTorch version 2.2.2 or later.

import torch
import torch.distributed.rpc as rpc

def function(code):
    """Vulnerable function that demonstrates the security issue."""
    return eval(code)

def start_rpc():
    """Vulnerable function that demonstrates the security issue."""
    rpc.init_rpc("worker", rank=0, world_size=1)
    rpc.register_function("function", function)

if __name__ == "__main__":
    start_rpc()
    result = rpc.rpc_sync("worker", "function", args=("print('Executing arbitrary code!')",))
    print(result)
    rpc.shutdown()

import torch
import torch.distributed.rpc as rpc
import ast

def function(code):
    """Secure function that fixes the vulnerability."""
    # SECURE: This version prevents command injection
    return ast.literal_eval(code)

def start_rpc():
    """Secure function that fixes the vulnerability."""
    rpc.init_rpc("worker", rank=0, world_size=1)
    rpc.register_function("function", function)

if __name__ == "__main__":
    start_rpc()
    result = rpc.rpc_sync("worker", "function", args=("print('Executing arbitrary code!')"))
    print(result)
    rpc.shutdown()

"__import__('os').system('cat /etc/passwd')"

@misc{vaitp:cve20245480,
  title        = {{RCE vulnerability in PyTorch RPC allows arbitrary command execution.}},
  author       = {Bogaerts, Fr\'ed\'eric and Ivaki, Naghmeh and Fonseca, Jos\'e},
  year         = {2024},
  note         = {VAITP Python Vulnerability Dataset, entry CVE-2024-5480},
  howpublished = {\url{https://netpack.pt/vaitp/vulnerability/CVE-2024-5480/}}
}