Security readout for executives and security teams
Plain-English summary
A flaw in vLLM can let an authorized user crash certain AI inference servers by submitting a malformed image to affected multimodal models. The business impact is service outage, not data theft. The issue is patched in vLLM 0.12.0.
Executive priority
Treat as a targeted availability risk for AI services. Prioritize patching internet-facing or broadly accessible multimodal vLLM deployments, especially production systems with uptime commitments.
Technical view
vLLM versions 0.6.4 through before 0.12.0 are affected when serving multimodal models using the Idefics3 vision model implementation. A specially crafted small image can trigger a tensor dimension mismatch and unhandled runtime error, terminating the server. CVSS is 6.5, with high availability impact.
Likely exposure
Exposure is likely where vLLM >=0.6.4 and <0.12.0 serves Idefics3-based multimodal endpoints to users. Text-only deployments, other model implementations, or vLLM 0.12.0+ are not identified as affected in the supplied sources.
Exploitation context
The CVE is not listed in KEV, and the provided sources do not report active exploitation. Exploitation requires network access and low privileges to submit image input to an affected endpoint. The documented impact is denial of service.
Researcher notes
Scope is narrow: vLLM Idefics3 vision model implementation, versions 0.6.4 to before 0.12.0. Avoid generalizing to all vLLM deployments without confirming model usage. The root cause is described as ambiguous image dimensions causing an unhandled runtime failure.
Mitigation direction
- Upgrade affected vLLM deployments to version 0.12.0 or later.
- Identify any Idefics3 multimodal serving configurations before prioritizing rollout.
- Limit access to multimodal inference endpoints to trusted users where possible.
- Review the vendor advisory for any updated guidance.
Validation and detection
- Inventory deployed vLLM versions across inference services.
- Confirm whether deployments serve Idefics3 vision-based multimodal models.
- Check service logs for tensor dimension mismatch errors and unexpected server termination.
- Verify upgraded services report vLLM 0.12.0 or later.
Public sources used
Generated from the cited source records. This long-tail analysis has not been individually reviewed by a named human.
Conservative CVE-to-ATT&CK context
These mappings and lookup hints may be relevant to the vulnerability behavior, CWE, affected product, or exposure path. Glexia-inferred context is not an official MITRE, ATT&CK, CWE, or CVE Program mapping.
ATT&CK lookup starting points
Use these exact CWE pages and searches to review the Glexia ATT&CK library from this CVE's weakness and description context.
CWE-770: Exact CWE lookup
Use the exact CWE identifier as the starting point before reviewing related ATT&CK behavior. Open the exact CWE lookup page first, then review the ATT&CK searches from that MITRE weakness context. This is a Glexia lookup hint, not an official ATT&CK mapping.
Open ATT&CK lookupCVE-2026-22773 mapping review
Open the CVE-to-ATT&CK bridge for reviewed, inferred, or future official mappings tied to this CVE.
Open ATT&CK lookup- Severity
- Medium
- CVSS
- 6.5 (3.1)
- Known Exploited
- No
- Published
Vector: CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H
CNA and ADP enrichment extracted from CVE v5
These fields come from the CVE record and ADP containers, not from Glexia's Take. They preserve time-varying source decisions such as CISA SSVC, KEV status, CVSS metrics, and provider references.
CVSS vector scores
1 official scoreWe collect every scored CVSS vector available in the official CNA and ADP containers. When more than one version is present, the table keeps the source vectors side by side instead of collapsing them into the highest score.
CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H2.83.6Primary CVE scoreVulnerability scoring details
Base CVSS 3.1 score
6.5MediumVector: CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H
Source materials
- CVE List V5 sourceCVE List V5
- https://github.com/vllm-project/vllm/security/advisories/GHSA-grg2-63fw-f2qrCVE reference · x_refsource_CONFIRM
Products and packages named in the record
CWE details
CWE links open Glexia weakness intelligence pages with official CWE context, developer remediation guidance, and related CVE mappings.
Allocation of Resources Without Limits or Throttling
Allocation of Resources Without Limits or Throttling represents a recurring weakness pattern that can create exploitable paths when design, validation, or implementation controls are missing.