VAITP Dataset

← Back to the dataset

CVE-2026-115912

Auth‑crafted shared‑memory data lets attackers inject pickle payload for RCE in vLLM <0.32.0.

  • CVSS 7.5
  • CWE-502
  • Input Validation and Sanitization
  • Remote

vLLM is an inference and serving engine for large language models. Prior to 0.32.0, an authenticated caller can place forged shared-memory address and monotonic_id values in public multimodal features.kwargs_data. In scale-out deployments with shared-memory multimodal caching enabled, workers trust those values and pass attacker-selected bytes to pickle.loads(), allowing arbitrary Python execution even when insecure serialization is disabled. This issue is fixed in version 0.32.0.

CVSS base score
7.5
Published
2026-10-26
OWASP
A08 Software and Data Integrity Failures
Orthogonal defect classification
Timing/Serialization
Code defect classification
Serialization Issues
Category
Input Validation and Sanitization
Subcategory
Insecure Parsing or Deserialization
Accessibility scope
Remote
Impact
Arbitrary Code Execution
Affected component
vLLM
Fixed by upgrading
Yes

Solution

Upgrade vLLM to version 0.32.0.

Vulnerable code sample

import pickle

def handle_multimodal_feature(kwargs_data):
    # Extract values that are assumed trustworthy in scale‑out deployments
    address = kwargs_data["shared_memory_address"]
    monotonic_id = kwargs_data["monotonic_id"]
    # Retrieve raw bytes from the shared‑memory cache (placeholder)
    raw_bytes = retrieve_from_shared_memory(address, monotonic_id)
    # VULNERABLE: unsafe deserialization of attacker‑controlled bytes
    feature = pickle.loads(raw_bytes)
    return feature

def retrieve_from_shared_memory(address, monotonic_id):
    # Placeholder for the actual shared‑memory fetch logic
    return b""

Patched code sample

import json

def handle_multimodal_feature(kwargs_data):
    # Extract values that are assumed trustworthy in scale‑out deployments
    address = kwargs_data["shared_memory_address"]
    monotonic_id = kwargs_data["monotonic_id"]
    # Retrieve raw bytes from the shared‑memory cache (placeholder)
    raw_bytes = retrieve_from_shared_memory(address, monotonic_id)
    # FIX: use safe deserialization instead of pickle
    feature = json.loads(raw_bytes.decode('utf-8'))
    return feature

def retrieve_from_shared_memory(address, monotonic_id):
    # Placeholder for the actual shared‑memory fetch logic
    return b""

Published fix

Public sources link this CVE to the changes below. The code samples above were generated to illustrate the defect; these links lead to the code the project actually changed.

  • [Security] Authenticate shared-memory multimodal cache handles (#59357) (vllm-project/vllm)
    vllm/distributed/device_communicators/shm_object_storage.py, vllm/distributed/device_communicators/shm_object_storage.py · SingleWriterShmObjectStorage.get, vllm/distributed/device_communicators/shm_object_storage.py · SingleWriterShmObjectStorage.touch, vllm/distributed/device_communicators/shm_object_storage.py · SingleWriterShmRingBuffer.access_buf, vllm/distributed/device_communicators/shm_object_storage.py · SingleWriterShmRingBuffer.allocate_buf, vllm/distributed/device_communicators/shm_object_storage.py · SingleWriterShmRingBuffer.clear, vllm/distributed/device_communicators/shm_object_storage.py · SingleWriterShmRingBuffer.__init__, vllm/entrypoints/scale_out/token_in_token_out/serving.py, vllm/entrypoints/scale_out/token_in_token_out/serving.py · ServingTokens.serve_tokens, vllm/multimodal/cache/base.py, vllm/multimodal/cache/shm.py, vllm/multimodal/cache/shm.py · ShmObjectStoreReceiverCache.get_and_update_item and 3 more

Cite this entry

@misc{vaitp:cve2026115912,
  title        = {{Auth‑crafted shared‑memory data lets attackers inject pickle payload for RCE in vLLM <0.32.0.}},
  author       = {Bogaerts, Fr\'ed\'eric and Ivaki, Naghmeh and Fonseca, Jos\'e},
  year         = {2026},
  note         = {VAITP Python Vulnerability Dataset, entry CVE-2026-115912},
  howpublished = {\url{https://netpack.pt/vaitp/vulnerability/CVE-2026-115912/}}
}

Download this entry as JSON Download the whole dataset

Introducing the "VAITP dataset": a specialized repository of Python vulnerabilities and patches, meticulously compiled for the use of the security research community. As Python's prominence grows, understanding and addressing potential security vulnerabilities become crucial. Crafted by and for the cybersecurity community, this dataset offers a valuable resource for researchers, analysts, and developers to analyze and mitigate the security risks associated with Python. Through the comprehensive exploration of vulnerabilities and corresponding patches, the VAITP dataset fosters a safer and more resilient Python ecosystem, encouraging collaborative advancements in programming security.

The supreme art of war is to subdue the enemy without fighting.

Sun Tzu – “The Art of War”

:: Shaping the future through research and ingenuity ::

Before you use the VAITP AI models

Please read and accept the following disclaimer:

By choosing "I understand and accept", you acknowledge that you understand and agree to these terms.