CVE-2026-93841: Improper Validation of Array Index in vllm-project vllm
vLLM through 0.29.0 contains a memory corruption vulnerability in the Triton _bincount_kernel where prompt token IDs index the penalty prompt-presence bitset without bounds checking against vocabulary size. Attackers can submit multimodal audio requests with tokens equal to vocabulary size, causing out-of-bounds writes that corrupt concurrent requests' sampler state and alter repetition penalty behavior.
AI Analysis
Technical Summary
CVE-2026-93841 describes a memory corruption vulnerability in vLLM through version 0.29.0. The flaw exists in the Triton _bincount_kernel where prompt token IDs are used to index the penalty prompt-presence bitset without verifying that the token IDs are within the vocabulary size bounds. Attackers can exploit this by submitting multimodal audio requests containing tokens equal to the vocabulary size, causing out-of-bounds writes. This memory corruption can alter the sampler state of concurrent requests, impacting the repetition penalty mechanism.
Potential Impact
The vulnerability allows out-of-bounds memory writes that corrupt the sampler state of concurrent requests, potentially altering the repetition penalty behavior. This could lead to unexpected or incorrect model outputs. The CVSS 4.0 score is 6.3 (medium severity), indicating a moderate impact with network attack vector and no privileges required.
Mitigation Recommendations
Patch status is not yet confirmed — check the vendor advisory for current remediation guidance. No official fix or patch links are provided in the available data. Until a patch is available, avoid processing untrusted multimodal audio requests with token IDs equal to or exceeding the vocabulary size.
CVE-2026-93841: Improper Validation of Array Index in vllm-project vllm
Description
vLLM through 0.29.0 contains a memory corruption vulnerability in the Triton _bincount_kernel where prompt token IDs index the penalty prompt-presence bitset without bounds checking against vocabulary size. Attackers can submit multimodal audio requests with tokens equal to vocabulary size, causing out-of-bounds writes that corrupt concurrent requests' sampler state and alter repetition penalty behavior.
CVSS v4.0
Score 6.3medium
Affected software
vllm-project
vllm
pkg:github/vllm-project/vllmRun on your own infrastructure? Check whether these packages are installed with threat-finder — our free open-source scanner.
AI-Powered Analysis
Machine-generated threat intelligence
Technical Analysis
CVE-2026-93841 describes a memory corruption vulnerability in vLLM through version 0.29.0. The flaw exists in the Triton _bincount_kernel where prompt token IDs are used to index the penalty prompt-presence bitset without verifying that the token IDs are within the vocabulary size bounds. Attackers can exploit this by submitting multimodal audio requests containing tokens equal to the vocabulary size, causing out-of-bounds writes. This memory corruption can alter the sampler state of concurrent requests, impacting the repetition penalty mechanism.
Potential Impact
The vulnerability allows out-of-bounds memory writes that corrupt the sampler state of concurrent requests, potentially altering the repetition penalty behavior. This could lead to unexpected or incorrect model outputs. The CVSS 4.0 score is 6.3 (medium severity), indicating a moderate impact with network attack vector and no privileges required.
Mitigation Recommendations
Patch status is not yet confirmed — check the vendor advisory for current remediation guidance. No official fix or patch links are provided in the available data. Until a patch is available, avoid processing untrusted multimodal audio requests with token IDs equal to or exceeding the vocabulary size.
Technical Details
- Data Version
- 5.2
- Assigner Short Name
- VulnCheck
- Date Reserved
- 2026-09-18T18:16:52.304Z
- Cvss Version
- 4.0
- State
- PUBLISHED
Threat ID: 6aad91cb55bf5e2cf5719192
Added to database: 09/18/2026, 19:32:27 UTC
Last enriched: 09/18/2026, 19:47:35 UTC
Last updated: 09/19/2026, 01:33:43 UTC
Views: 4
Community Reviews
0 reviewsCrowdsource mitigation strategies, share intel context, and vote on the most helpful responses. Sign in to add your voice and help keep defenders ahead.
Want to contribute mitigation steps or threat intel context? Sign in or create an account to join the community discussion.
Actions
Updates to AI analysis require Pro Console access. Upgrade inside Console → Billing.
Need more coverage?
Upgrade to Pro Console for AI refresh and higher limits.
For incident response and remediation, OffSeq services can help resolve threats faster.
Latest Threats
Check if your credentials are on the dark web
Instant breach scanning across billions of leaked records. Free tier available.