Live Exposure Defense: From CVE to Confirmed Exposure in 12 Hours – See more

Back
New CVE Detected

CVE-2026-94627 – GPU Memory Exhaustion (DoS) – vLLM Mooncake Connector ≤ 0.29.0

Be the first to know when new zero-days emerge:

Subscribe to Email Alerts
Subscribe to RSS Feed

Summary

CVE-2026-94627 is a denial-of-service vulnerability in vLLM’s Mooncake KV-transfer connector, used in prefill/decode disaggregated serving deployments. The connector fails to properly manage GPU KV cache block ownership when concurrent child requests share a single transfer ID, allowing an unauthenticated remote attacker to exhaust GPU memory and degrade or disable the inference service. The issue is rated HIGH severity and affects all vLLM releases through 0.29.0 that use the Mooncake connector.

Technical details

  • Root cause: In mooncake_connector.py, when a completion request containing multiple prompts spawns concurrent "child" requests, those child requests can share a single Mooncake transfer ID. The connector does not correctly track per-request ownership of GPU KV cache blocks tied to that shared transfer ID.
  • Trigger conditions: The target vLLM instance must be running in a prefill/decode (P/D) disaggregated deployment using the Mooncake KV cache transfer connector. An attacker submits completion requests containing multiple prompts, causing overlapping/duplicate transfer IDs.
  • Attack vector: Network-based; no authentication or user interaction is required to send the crafted completion requests.
  • Impact: Because ownership of the shared transfer ID is mismanaged, KV cache blocks become orphaned rather than released when a request completes. These orphaned blocks accumulate in GPU memory over repeated requests until the process is restarted, ultimately exhausting GPU memory and blocking legitimate inference requests (availability impact only — no confidentiality or integrity impact).

Affected software

  • vllm-project vLLM — all versions through 0.29.0 that are deployed with the Mooncake connector in prefill/decode disaggregated configurations (vllm/distributed/kv_transfer/kv_connector/v1/mooncake/mooncake_connector.py)

Severity

  • CVSS v3.1 Base Score: 7.5 (HIGH) — CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H
  • CVSS v4.0 Base Score: 8.7 (HIGH) — CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N
  • CWE-401: Missing Release of Memory after Effective Lifetime

Mitigation and recommended actions

  • Immediate: Upgrade to a vLLM release that incorporates the fix from pull request #49796, which rejects duplicate/reused Mooncake transfer_id values to prevent KV cache leakage. Confirm the running version is newer than 0.29.0 and includes this fix before continuing to use the Mooncake connector in production.
  • If immediate patching is not possible:
    • Disable or avoid the Mooncake KV-transfer connector in prefill/decode disaggregated deployments until patched.
    • Restrict network access to the vLLM API endpoint to trusted clients only (e.g., via firewall rules, API gateway authentication, or network segmentation), since the flaw requires no authentication to trigger.
    • Monitor GPU memory utilization and implement automated process restarts/alerting to reduce the impact of cache-block accumulation until the patch can be applied.
    • Rate-limit or restrict completion requests containing large numbers of prompts, since these are the trigger for the orphaned-block accumulation.

References

Are you exposed?

Get a free report of your organization’s exposure to this CVE and threat

  • Mapping of all Assets with this Technology
  • Identification of potentially exposed assets to this CVE
  • Confirmation of verified exploitable assets
Get Exposure Report

How IONIX’s External Exposure Management Platform Detects and Validates
Zero-Days to Shrink MTTR

1

Map your entire attack surface (continously)

IONIX uses multi-factor discovery methods, including DNS analysis, certificate mapping, metadata inspection, and more, to automatically map every internet-facing asset across your environment. This includes cloud instances, third-party platforms, shadow IT, and even forgotten infrastructure that traditional tools miss.

2

Monitor for new CVEs

Dozens of threat intel feeds using agentic technology are continuously analyzed to detect the appearance of proof-of-concept code, exploit kits, and indicators of active targeting. IONIX goes further by applying AI to proactively evaluate whether emerging vulnerabilities are likely to be exploited, even before PoCs go public.

3

Identify Potential External Exposures

Not all CVEs matter. IONIX filters vulnerabilities by asking attacker-centric questions: Can it be reached from the internet? Does it require authentication? Is it being exploited in the wild? This dramatically reduces noise and focuses teams on threats that can actually be weaponized.

4

Create Safe, Scalable Exploit Validations

IONIX transforms real-world PoCs into safe, non-intrusive test payloads that can be run in production environments without disruption. These simulations are precisely targeted to the systems that are vulnerable, ensuring rapid validation without unnecessary load.

5

Execute Exploit Validations

By combining context about software stack, versioning, exposure status, and reachability, IONIX ensures that only the right payloads are executed against the right assets, maximizing efficiency and minimizing risk.

6

Drive Fast and Actionable Remediation

Results are routed through integrations with ticketing, SOAR, and SIEM tools. Issues are written in plain language, bundled into remediation clusters, and prioritized based on asset criticality, exploitability, and blast radius. This shortens mean time to remediation (MTTR) and empowers teams to act with confidence.

Are you exposed?

Get a free report of your organization’s exposure to this CVE and threat

  • Mapping of all Assets with this Technology
  • Identification of potentially exposed assets to this CVE
  • Confirmation of verified exploitable assets
Get Exposure Report
Close

Subscribe to Threat Center RSS

Copy/paste the link below into your preferred RSS reader or follow these instructions to subscribe to Slack alerts.

Open RSS feedSlack setup instructions

Close

Get Real-Time CVE Alerts to Your Email

Be the first to know when new zero-days emerge