High Memory corruption

CVE-2026-94627 — vLLM through 0.29.0 GPU KV Cache Leak via Mooncake Transfer ID Collision

vLLM Mooncake connector through 0.29.0 fails to properly manage GPU KV cache block ownership when concurrent child requests share a single transfer ID in prefill/decode disaggregated deployments. Attackers can trigger GPU memory exhaustion by submitting completion requests with multiple prompts, causing orphaned KV cache blocks to accumulate until process restart and eventually preventing legitimate requests from executing.

Published Sep 21, 2026 · Updated Oct 15, 2026 · Sources: NIST National Vulnerability Database, CVE Program (cve.org) e CISA ADP · CNA: VulnCheck

Low

Who exploits it and how urgent it is

A 0–100 score combining confirmed exploitation, ransomware and APT use, public exploits, EPSS and severity.

No exploitation signal in the sources consulted so far.

How the score was calculated

EPSS 0.6% (30-day exploitation probability)+7
Technical severity (high)+6
Automatable exploitation (SSVC)+3

The score orders the patch queue; it does not replace your asset context (internet exposure, criticality, compensating controls).

Triage

Is it exploited, how likely is exploitation, what does it touch, and how severe do the sources call it.

Exploitation

Not confirmed

Not in CISA KEV — an absence of record, not proof it is unexploited.

EPSS FIRST

0.63%

ahead of 49% of scored CVEs · Oct 15, 2026

Affects

Vllm

Vllm and 1 more

CVSS base

7.5

High · 2 source(s)

CISA SSVC assessment CISA

Three decision points CISA publishes for the CVEs it assesses (SSVC 2.0.3). A stakeholder decision, not a severity score.

Exploitation

None

none · proof-of-concept · active

Automatable

Yes

can an attacker script the kill-chain steps

Technical impact

Partial

partial · total control of the vulnerable component

Remediation

Fixes and mitigations documented by the sources — linked to the original advisory.

Documented patch — github.com (NIST National Vulnerability Database)

First 24 hours

Ordered from the record’s own fields — exposure first, because you cannot patch what you have not found. A rule-generated starting point.

Identify exposed assets running the affected vendor/product/version combinations (Vllm Vllm, Vllm-project Vllm-project/vllm).
The attack is network-based and needs no privileges: restrict exposure (ACLs, VPN, WAF) until patched.
Apply the fixed version or the patch documented under Remediation.
Search available logs for exploit probes matching the vulnerability class (memory corruption), errors and authentication anomalies.

Affected scope

Vendor, product and versions as the sources word them.

Vendor	Product	Versions	Status	Source
Vllm	Vllm	≤ 0.29.0	Vulnerable	NIST National Vulnerability Database
Vllm-project	Vllm-project/vllm	≤ 0.29.0	Vulnerable	CVE Program (cve.org) e CISA ADP

Attack characteristics

The CVSS vector, decoded. It describes the attack, not your exposure to it.

Attack vector

Network

Attack complexity

Low

Privileges required

None

User interaction

None

Scope

Unchanged

Confidentiality

None

Integrity

None

Availability

High

Every base score collected

Sources score independently and may disagree; each row says who scored it and under which version.

Score	Version	Severity	Expl.	Impact	Scored by
8.7	CVSS 4.0	High	—	—	disclosure@vulncheck.com via NIST National Vulnerability Database
8.7	CVSS 4.0	High	—	—	VulnCheck via CVE Program (cve.org) e CISA ADP
7.5	CVSS 3.1	High	3.9	3.6	disclosure@vulncheck.com via NIST National Vulnerability Database
7.5	CVSS 3.1	High	—	—	VulnCheck via CVE Program (cve.org) e CISA ADP

CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H (score shown at the top: disclosure@vulncheck.com)

Weakness MITRE

Weakness class according to MITRE’s CWE catalog.

CWE-401 · Missing Release of Memory after Effective Lifetime

Other CVEs with this weakness →

Detection

Read off the CVSS vector and the weakness class. Starting points, not tested rules.

Prioritize edge telemetry for the network-reachable component (proxy, WAF, IDS/IPS).
Correlate service crashes/restarts with inbound traffic; repeated crashes may indicate attempts.

web/proxy/WAF logsnetwork IDS/IPSendpoint telemetry (EDR)crash dumps

Timeline

What happened to this CVE, newest first.

2026
Affected products added · vllm-project/vllm
Oct 15, 2026 · CVE Program (cve.org) e CISA ADP
SSVC assessment updated
Oct 15, 2026 · CVE Program (cve.org) e CISA ADP
Record published
Oct 15, 2026
Fix documented
Oct 15, 2026 · NIST National Vulnerability Database
CVSS changed · 7.5
Oct 15, 2026 · NIST National Vulnerability Database
Source status changed · → Analyzed
Oct 15, 2026 · NIST National Vulnerability Database
Record created in this database
Oct 15, 2026 · NIST National Vulnerability Database
CVE published
Sep 21, 2026 · VulnCheck

Source activity

Repeated readings are counted rather than listed.

1× Official description updated · Oct 15, 2026 – Oct 15, 2026

References

4 on the record

github.com/vllm-project/vllm/pull/49796
Issue Tracking, Patch, issue-tracking, patch, NIST National Vulnerability Database
github.com/vllm-project/vllm/blob/v0.29.0/vllm/distributed/kv_transfer/kv_connec
Product, technical-description, NIST National Vulnerability Database
www.vulncheck.com/advisories/vllm-through-0.29.0-gpu-kv-cache-leak-via-mooncake-transfe
Third Party Advisory, third-party-advisory, NIST National Vulnerability Database
github.com/vllm-project/vllm
Product, product, NIST National Vulnerability Database

Elsewhere on this site

Vllm
every CVE for this vendor

Memory corruption
same class

CWE-401
other pages naming this weakness

Related CVEs (same products)

Medium CVE-2026-109311 vLLM before 0.29.0 Heap Overflow DoS via IAMF Audio Transcription Upload 3d
Medium CVE-2026-105922 vllm-project vLLM Penalty utils.py get_token_bin_counts_and_mask denial of service 9d
Medium CVE-2026-105775 vllm-project vLLM Completions Request mamba_mixer2.py conv_ssm_forward out-of-bounds 10d
Medium CVE-2026-105756 vLLM: Loose `cache_salt` validation lets a single request kill EngineCore on LMCache-MP deployments — uncaught downstream `ValueError` denial of service 10d
Medium CVE-2026-105757 vLLM: Structured-output request errors escape the request boundary and terminate the shared EngineCore — engine-fatal denial of service (3 sites) 10d
Medium CVE-2026-105755 vLLM: Flash late-interaction scoring caches query embeddings under a caller-controlled request id — cross-request integrity break and induced errors on `/score` and `/rerank` 10d

Not in any source we poll

Listed rather than left blank: an empty field and an unmeasured one look identical on screen, and only one is a reason to look elsewhere.

No confirmed IOCs, IP addresses, domains, file hashes, or malware artifacts supplied.
No organization-specific asset inventory, compensating-control status, or patch deployment evidence supplied.
No exploit packet captures, log samples, or incident case IDs supplied.

Answered from this record

Text assembled only from the fields above — no additional claims.

What should defenders know first?

CVE-2026-94627 is "vLLM through 0.29.0 GPU KV Cache Leak via Mooncake Transfer ID Collision", rated High. It affects Vllm (Vllm); versions are in the scope table. It is not in CISA KEV, which does not rule out exploitation. CVSS 7.5 describes technical severity. EPSS 0.63% estimates the probability of exploitation in 30 days; neither replaces your asset context. First verify whether exposed or business-critical assets run the affected versions, apply vendor fixes or mitigations, and preserve logs for review.

Comments

Moderated technical discussion. Read the community guidelines.

No comments yet.