Improper Validation of Syntactic Correctness of Input Affecting vllm package, versions [,0.5.5)


Severity

Recommended
0.0
high
0
10

CVSS assessment made by Snyk's Security Team

    Threat Intelligence

    Exploit Maturity
    Proof of concept
    EPSS
    0.05% (17th percentile)

Do your applications use this vulnerable package?

In a few clicks we can analyze your entire application and see what components are vulnerable in your application, and suggest you quick fixes.

Test your applications
  • Snyk ID SNYK-PYTHON-VLLM-7981556
  • published 15 Sep 2024
  • disclosed 22 Aug 2024
  • credit Unknown

How to fix?

Upgrade vllm to version 0.5.5 or higher.

Overview

vllm is an A high-throughput and memory-efficient inference and serving engine for LLMs

Affected versions of this package are vulnerable to Improper Validation of Syntactic Correctness of Input in the process_model_inputs() and process_model_inputs_async() functions, accessible through the completions API. An attacker can crash the server by sending a request with and empty prompt, if a model that does not prepend any data (such as gpt2) is in use by the server. The crash will only happen if the processed prompt that is passed to these functions is still empty.

PoC

{
     "model": "gpt2",
     "prompt": [""],
     "max_tokens": 20,
     "temperature": 0
}

CVSS Scores

version 4.0
version 3.1
Expand this section

Snyk

Recommended
8.2 high
  • Attack Vector (AV)
    Network
  • Attack Complexity (AC)
    Low
  • Attack Requirements (AT)
    Present
  • Privileges Required (PR)
    None
  • User Interaction (UI)
    None
  • Confidentiality (VC)
    None
  • Integrity (VI)
    None
  • Availability (VA)
    High
  • Confidentiality (SC)
    None
  • Integrity (SI)
    None
  • Availability (SA)
    None