CVE-2026-105755

What it is

vLLM is an inference and serving engine for large language models. Prior to 0.30.0, flash late-interaction scoring at the /score and /rerank endpoints derives each worker's query_key value from the caller-controlled X-Request-Id header. A concurrent request that reuses a victim's identifier can overwrite the cached query embedding so the victim's documents are scored against the attacker's query, and shared use counters can also cause a late-interaction cache-miss error. This issue is fixed in version 0.30.0.

At a glance

Exploitability

Can be exploited over the network with a low-privilege account. Limited direct impact.

Exploit odds

0.2% chance of exploitation in the next 30 days (EPSS).

Fix

A patch, advisory or fixed version is referenced. Upgrade or apply it.

Timeline

1. 2026-10-05 23:17 UTC PublishedDisclosed publicly on CVE.org and NVD
2. 2026-10-08 01:28 UTC Last updatedMost recent change to the record

Why it ranks here

• Exploited in the wild: no, CISA does not list it.
• Exploit odds: 0.2% in the next 30 days (EPSS).
• Severity: 4.2 MEDIUM (CVSS).
• Fix: a patch or advisory is linked.

Details

Published

2026-10-05 23:17 UTC

Last modified

2026-10-08 01:28 UTC

Exploit probability

0.2% · EPSS, next 30 days

Weakness

CWE-639 · Insecure direct object reference

Record status

Analyzed

Affected

• Vllm < 0.30.0

Attack profile CVSS 3.1 · 4.2

Attack vector

Network

Complexity

High

Privileges needed

Low

User interaction

None

Scope

Unchanged

Confidentiality

None

Integrity

Low

Availability

Low

Is it in your code?

Hand this to your coding agent. It searches your project for the affected products and checks versions. Nothing leaves your machine.

Read https://cve-list.com/agent/cve/CVE-2026-105755.md and follow it. Check whether this project is affected by CVE-2026-105755. Read only. Don't change anything and don't upload my code. Tell me yes, no or not sure, with the file as proof.

Copy as ticket · More options · Status badge