CVE-2026-105922: vllm-project vLLM Penalty utils.py get_token_bin_counts_and_mask denial of service
A security flaw has been discovered in vllm-project vLLM up to 0.31.0. This impacts the function gettokenbincountsandmask of the file vllm/modelexecutor/layers/utils.py of the component Penalty Handler. Performing a manipulation results in denial of service. Remote exploitation of the attack is possible. The exploit has been released to the public and may be used for attacks. The project was informed of the problem early through an issue report but has not responded yet.
Affected Software
Event History
Frequently Asked Questions
Which deployments are affected?
vLLM versions up to and including 0.31.0 are reported as affected, specifically the Penalty Handler code in vllm/model_executor/layers/utils.py.
What access does an attacker need?
The issue is remotely exploitable and requires low privileges. No user interaction is required.
What is the practical impact?
Successful exploitation can cause a denial of service. The provided information does not indicate confidentiality or integrity impact.
Is exploitation likely?
A public exploit has been released and may be used in attacks. The exploitability assessment is rated as proven.
Is a fix available?
The provided information states that the project had been notified through an issue report but had not responded; it does not identify a fixed version or workaround.