GHSA-7m6h-x95x-82q5 is a medium-severity (CVSS 5.3) CWE-190 vulnerability in vllm. A fix is available for vllm — see the affected versions and patch details below.
Exploitation Status
Proof-of-concept exploit code exists
- CISA’s SSVC triage found public proof-of-concept exploit code for this CVE, though no confirmed active exploitation.
Exploitation and automatability from CISA’s SSVC triage for GHSA-7m6h-x95x-82q5.
EPSS Exploitation Probability
Probability of exploitation in the next 30 days, from FIRST.org EPSS.
How urgent is this, really
GHSA-7m6h-x95x-82q5 by exploitation likelihood (EPSS) against impact (CVSS). Outside the shaded patch-first corner.
Where this sits among everything scored
Of 395,249 CVEs with a current EPSS score, this one falls in the < 10% band (highlighted). Counts from FIRST.org, log-scaled.
Real-World Exposure
vllmReal-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
An integer overflow in the act_and_mul_kernel kernel can cause the output of one user request to be incorporated into the response of another request within the same inference batch. Under certain conditions, the last request in a batch can receive a partial or complete copy of the first user's inference result, resulting in cross-user data leakage.
Details
The root cause is an integer overflow in the expression blockIdx.x * 2 * d at https://github.com/vllm-project/vllm/blob/ff712f6447093d07747c88680b9d006b119f5890/csrc/activation_kernels.cu#L82.
As a result, the computation for one user (User A) can incorrectly consume input data from another user (User B). In particular, when 2^32 is divisible by d, the overflow can cause User A's output to contain portions of User B's inference result. In some cases, User B's response may be copied entirely into User A's response.
This constitutes a severe cross-user information disclosure vulnerability and is straightforward to trigger.
PoC
We reproduced the issue using meta-llama/Llama-3.2-1B-Instruct, for which d = 8192.
Using the following configuration:
Batch size: 17 Sequence length: 16384
The final response in the batch becomes an exact copy of the first response in the batch, demonstrating complete cross-user data leakage.
Impact
This vulnerability enables cross-user information disclosure. An attacker can intentionally craft requests that are processed within the same inference batch as a victim's request and cause the victim's inference output to be copied into the attacker's response.
As a result, sensitive information contained in another user's model response may be exposed to an unauthorized party.
Versions
For versions prior and equal to 0.21.0, the bug is in csrc/activation_kernels.cu, and for versions later than 0.21.0, the bug is in csrc/libtorch_stable/activation_kernels.cu.
Affected Packages
| Ecosystem | Package | Vulnerable range | Fix |
|---|---|---|---|
| 🐍PyPI | vllm | all versions | 0.27.0pip install --upgrade 'vllm==0.27.0' |
Affected Products
vllmvllmDetection & mitigation playbook
Open-source dependencyDetect
Scan your dependency tree (package-lock.json, pnpm-lock.yaml, requirements.txt, go.sum, etc.) for vllm, including transitive dependencies — a direct dependency you never call can still pull in a vulnerable version.
Fix
Update vllm to 0.27.0 or later, then make sure no transitive (indirect) dependency still pins the vulnerable range — O3 confirms GHSA-7m6h-x95x-82q5 is resolved across your whole dependency graph.
Workarounds
Constrain what reaches the vulnerable code: limit the size and shape of untrusted input, isolate the affected component in a sandboxed or least-privileged process, and enable the platform's memory-safety mitigations (ASLR, stack protector, hardened allocator) so an out-of-bounds access is more likely to fail closed than to be exploitable.
Fixing This On Your OS
If you run this on a Linux distribution, patch through your package manager against the distro's own security advisory below — it tracks the exact backported fix for your release, which can ship on a different timeline (and sometimes a different severity) than the upstream project.
This Moderate flaw in vLLM, an inference engine for large language models, can lead to cross-user data leakage. An integer overflow during batched inference may allow one user's request to inadvertently receive partial or complete inference results from another user within the same batch. This is particularly relevant…
| Product | Fixed in | Advisory |
|---|---|---|
| Red Hat AI Inference Server 3.5 | rhaii/vllm-rocm-rhel9:1791425869 | RHSA-2026:82059 |
| Red Hat AI Inference Server 3.5 | rhaii/vllm-gaudi-rhel9:1791427182 | RHSA-2026:82061 |
| Red Hat AI Inference Server 3.5 | rhaii/vllm-cuda-rhel9:1791473631 | RHSA-2026:82368 |
| Red Hat AI Inference Server 3.5 | rhaii/vllm-spyre-rhel9:1791818357 | RHSA-2026:82371 |
| Red Hat AI Inference Server 3.5 | rhaii/vllm-cpu-rhel9:1791473614 | RHSA-2026:82415 |
Frequently Asked Questions
Is GHSA-7m6h-x95x-82q5 in your dependencies?
Find it across PyPI, including transitive dependencies.