PHP2Shell : See if you’re affected with a free external attack surface scan. Click Here > 

Back
New CVE Detected

CVE-2026-94622 – Denial of Service (Uncaught KeyError) – vLLM NIXL Connector (≤ 0.29.0)

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

Subscribe to Email Alerts
Subscribe to RSS Feed

Summary

CVE-2026-94622 is a denial-of-service vulnerability affecting vLLM, the widely used open-source LLM inference and serving engine, in versions up to and including 0.29.0. The flaw resides in the NIXL KV-transfer connector used for prefill/decode disaggregated deployments, where malformed client-supplied requests can trigger an uncaught exception that crashes the decode engine. The issue carries a CVSS v3.1 base score of 7.5 (High) and a CVSS v4.0 score of 8.7 (High), and is remotely exploitable without authentication or user interaction.

Technical details

  • Root cause: The NIXL connector’s metadata handling (vllm/distributed/kv_transfer/kv_connector/v1/nixl/metadata.py) does not validate that all required fields are present in the client-supplied kv_transfer_params dictionary before indexing into it.
  • Trigger conditions: An attacker sends a request with an incomplete kv_transfer_params payload (e.g., setting do_remote_prefill: true while omitting required fields such as remote_block_ids, remote_engine_id, remote_request_id, remote_host, or remote_port).
  • Attack vector: Network-based; exploitable by any client able to reach a vLLM deployment configured for prefill/decode disaggregation using the NIXL KV connector. No authentication or special privileges are required (CWE-248: Uncaught Exception).
  • Impact: The missing keys cause an uncaught KeyError during EngineCore scheduling, which crashes the decode engine process. This halts all in-flight and subsequently routed inference requests until an operator manually restarts the engine — a full availability (denial-of-service) impact with no confidentiality or integrity loss.
  • Exposure: Organizations running distributed, high-throughput vLLM clusters with NIXL KV transfer enabled for disaggregated prefill/decode serving — including internet-facing inference gateways, multi-tenant LLM platforms, and GPU-as-a-service environments — are most exposed.

Affected software

  • vllm-project/vllm — all versions through and including 0.29.0 that use the NIXL KV connector (v1/nixl) for disaggregated prefill/decode deployments.

Severity

  • CVSS v3.1 Base Score: 7.5 (High)
  • CVSS v3.1 Vector: CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H
  • CVSS v4.0 Score: 8.7 (High)
  • CWE-248: Uncaught Exception

Mitigation and recommended actions

  • Immediate: Upgrade vLLM to a release that incorporates the fix from pull request #54807, which adds validation of kv_transfer_params before NIXL receive registration (checking for all required remote fields and range-validating remote_block_ids prior to use). Deployments should track the vLLM GitHub repository/release notes for the first tagged release after 0.29.0 that includes this patch and update as soon as it is available.
  • If immediate patching is not possible:
    • Restrict network access to disaggregated prefill/decode endpoints (e.g., NIXL connector ports) to trusted internal services only; do not expose them directly to untrusted clients or the public internet.
    • Deploy a validating proxy or gateway in front of the vLLM API that rejects requests containing incomplete or malformed kv_transfer_params payloads.
    • Monitor decode-engine processes for unexpected crashes/restarts and set up automated recovery/alerting to reduce downtime until the patch is applied.

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