Your RSA-2048 keys break in 2030. Find every one of them before attackers do.
🐍
🐍 PyPI
Not in CISA KEV
MEDIUM severity

GHSA-83vm-p52w-f9pw

MEDIUMFix: vllm-project/vllm#38610

GHSA-83vm-p52w-f9pw is a medium-severity (CVSS 6.5) CWE-131 vulnerability in vllm. O3 Security confirms whether GHSA-83vm-p52w-f9pw is actually reachable in your code before you act, and blocks exploitation at runtime until you patch.

vLLM: extract_hidden_states speculative decoding crashes server on any request with penalty parameters

Also known asCVE-2026-44223PYSEC-2026-145
Published
May 6, 2026
Updated
Jun 8, 2026
Affected
1 pkg
Patched
1 / 1
Exploits
None indexed
Exploitation data as of Jun 8, 2026 · OSV.dev, NVD, FIRST.org (EPSS)

Real-World Exposure

1 pkg affected
🐍vllm

Real-time download stats are indexed for npm and PyPI packages. This vulnerability affects PyPI packages — download data is not available via public APIs for these ecosystems.

Description

Summary

The extract_hidden_states speculative decoding proposer in vLLM returns a tensor with an incorrect shape after the first decode step, causing a RuntimeError that crashes the EngineCore process. The crash is triggered when any request in the batch uses sampling penalty parameters (repetition_penalty, frequency_penalty, or presence_penalty).

A single request with a penalty parameter (e.g., "repetition_penalty": 1.1) is sufficient to crash the server. The crash is deterministic and immediate — no concurrency, race condition, or special workload is required.

Details

In vLLM v0.17.0, the extract_hidden_states proposer's propose() method returned sampled_token_ids.unsqueeze(-1), producing a tensor of shape (batch_size, 1).

In PR #37013 (first released in v0.18.0), the KV connector interface was refactored out of propose(). The return type changed from tuple[Tensor, KVConnectorOutput | None] to Tensor, and the .unsqueeze(-1) call was removed along with the KV connector output:

# Before (v0.17.0):
return sampled_token_ids.unsqueeze(-1), kv_connector_output  # shape (batch_size, 1)

# After (v0.18.0+):
return sampled_token_ids  # shape (batch_size, 2) after first decode step

The refactor missed that sampled_token_ids changed semantics between the first and subsequent decode steps. After the first decode step, the rejection sampler allocates its output as (batch_size, max_spec_len + 1). With num_speculative_tokens=1, this produces shape (batch_size, 2) instead of the expected (batch_size, 1), causing a broadcast shape mismatch during penalty application.

Impact

Any vLLM deployment between v0.18.0 and v0.19.1 (inclusive) configured with extract_hidden_states speculative decoding is affected. A single API request containing any penalty parameter immediately and permanently crashes the EngineCore process, resulting in complete loss of service availability.

Patches

Fixed in PR #38610, first included in vLLM v0.20.0. The fix slices the return value to sampled_token_ids[:, :1], ensuring the correct (batch_size, 1) shape regardless of the rejection sampler's output dimensions.

Workarounds

  • Upgrade to vLLM v0.20.0 or later.
  • If upgrading is not possible, avoid using extract_hidden_states as the speculative decoding method on affected versions.
  • Alternatively, reject or strip penalty parameters (repetition_penalty, frequency_penalty, presence_penalty) from incoming requests at an API gateway before they reach vLLM.

Affected Packages

1 total 1 fixed
EcosystemPackageVulnerable rangeFix
🐍PyPIvllm0.18.0&&< 0.20.00.20.0

Detection & mitigation playbook

Open-source dependency
  1. Detect

    Scan your dependency tree (package-lock.json, pnpm-lock.yaml, requirements.txt, go.sum, etc.) for vllm. O3's reachability analysis confirms whether the vulnerable code path is actually invoked in your application, so you act on real exposure instead of every transitive match.

  2. Fix

    Update vllm to 0.20.0 or later, then make sure no transitive (indirect) dependency still pins the vulnerable range — O3 confirms GHSA-83vm-p52w-f9pw is resolved across your whole dependency graph.

  3. Workarounds

    If you can't upgrade right away: gate or disable the affected feature, validate untrusted input at the boundary, and avoid passing attacker-controlled data into the vulnerable path. O3's runtime protection blocks exploitation in production as an interim safeguard until the upgrade lands.

  4. How O3 protects you

    O3 pinpoints whether GHSA-83vm-p52w-f9pw is reachable in your code and exactly where to fix it, then blocks exploitation in production at runtime until the patched version is deployed.

Tailored to GHSA-83vm-p52w-f9pw. Runtime protection reduces exposure until a permanent patch is applied and verified — it complements patching, it doesn't replace it.

Frequently Asked Questions

### Summary The `extract_hidden_states` speculative decoding proposer in vLLM returns a tensor with an incorrect shape after the first decode step, causing a `RuntimeError` that crashes the EngineCore process. The crash is triggered when any request in the batch uses sampling penalty parameters (`repetition_penalty`, `frequency_penalty`, or `presence_penalty`). A single request with a penalty parameter (e.g., `"repetition_penalty": 1.1`) is sufficient to crash the server. The crash is deterministic and immediate — no concurrency, race condition, or special workload is required. ### Details
O3 Security · Impact-Aware SCA

Is GHSA-83vm-p52w-f9pw in your dependencies?

O3 detects GHSA-83vm-p52w-f9pw across PyPI dependencies and uses function-level reachability to confirm whether the vulnerable code path is actually reachable — not just present. No false positives.

GHSA-83vm-p52w-f9pw: vllm (Medium 6.5) | O3 Security