CVE-2026-105752: vllm Information Disclosure — Fix

Description

vLLM is an inference and serving engine for large language models. Prior to 0.30.0, Harmony tool continuations submitted through "POST /v1/responses" requests rebuild the next-turn engine input without preserving the cache_salt value, placing the continuation prefix in the global unsalted cache namespace even when the caller enabled salting. On deployments with prefix caching enabled, which is the default, an authenticated tenant who can reconstruct a victim's low-entropy post-tool history can submit the same continuation and use the cached_tokens_per_turn count to determine whether the prefix was previously processed, defeating the intended tenant isolation of salted prefix caching. This issue is fixed in version 0.30.0.

Is your site exposed to CVE-2026-105752?

Run a free security scan — no signup, results in seconds.

CVSS v3.1 Score

3.1

LOW

CVSS:3.1/AV:N/AC:H/PR:L/UI:N/S:U/C:N/I:L/A:N

EPSS — Exploit Prediction

0.0021

Probability of exploitation

0.11%

Percentile rank

EPSS estimates the probability that this vulnerability will be exploited in the wild within the next 30 days. A higher score means more likely to be exploited.

Weakness Type (CWE)

CWE-200 CWE-200

CWE-524 CWE-524

Affected Products

Vendor	Product	Versions
vllm	vllm	< 0.30.0

References

Advisories & Patches

https://github.com/vllm-project/vllm/commit/6a2a2bb02b563b83f946012959fd3927984d072a https://github.com/vllm-project/vllm/pull/50195 https://github.com/vllm-project/vllm/pull/51818 https://github.com/vllm-project/vllm/security/advisories/GHSA-935w-9g4m-p28p

Other References

https://github.com/vllm-project/vllm/releases/tag/v0.30.0

Frequently Asked Questions

What is CVE-2026-105752? +

vLLM is an inference and serving engine for large language models. Prior to 0.30.0, Harmony tool continuations submitted through "POST /v1/responses" requests rebuild the next-turn engine input without preserving the cache_salt value, placing the continuation prefix in the global unsalted cache namespace even when the caller enabled salting. On deployments with prefix caching enabled, which is the default, an authenticated tenant who can reconstruct a victim's low-entropy post-tool history can submit the same continuation and use the cached_tokens_per_turn count to determine whether the prefix was previously processed, defeating the intended tenant isolation of salted prefix caching. This issue is fixed in version 0.30.0. It has a CVSS v3.1 base score of 3.1 (LOW).

How severe is CVE-2026-105752? +

CVE-2026-105752 has a CVSS v3.1 score of 3.1 out of 10, rated LOW. This is a low-severity vulnerability with limited impact. The EPSS score is 0.0021, placing it in the 0th percentile for exploitation probability.

What products are affected by CVE-2026-105752? +

CVE-2026-105752 affects products from vllm, specifically: vllm. Check the affected products table above for specific version ranges.

How do I check if I'm vulnerable to CVE-2026-105752? +

You can use Secably's free Website Scanner to check your website for known vulnerabilities. For infrastructure scanning, use the Port Scanner to identify exposed services that may be affected. Check the vendor advisories linked above for specific patch and version information.

Related Vulnerabilities

CVE-2025-15625

Unauthenticated user is able to execute arbitrary SQL commands in Sparx Pro Cloud Server database in certain cases.

CVE-2026-22051

StorageGRID (formerly StorageGRID Webscale) versions prior to 11.9.0.13 and 12.0.0.6 are susceptible to a Information Disclosure vulnerability. Successful exploit could …

CVE-2026-41186

When Calico's shared debug server is enabled (disabled by default), the Calico kube-controllers and Goldmane components bind their Go pprof …

CVE-2026-54396

An information disclosure vulnerability exists in the MISP AuthKey edit functionality. When a validation error occurs during an AuthKey edit …

CVE-2026-13230

An information disclosure vulnerability was identified in TP-Link Kasa EC70 v4 and EC71 v4 in the local discovery mechanism, which …

CVE-2026-53912

Cerebrate before version 1.37 exposed credential material from self-registration requests. The self-registration workflow stored the registrant’s hashed password in the …

CVE-2026-105752