01 Record
CVE-2026-115913
vLLM: Phi-4 multimodal audio encoder materializes and retains full attention matrices: a single max-length audio OOM-crashes the vLLM engine
MediumPUBLISHEDCNA: GitHub_M
CNA base score6.5 CVSS v3.xCVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H
02 Description
vLLM: Phi-4 multimodal audio encoder materializes and retains full attention matrices: a single max-length audio OOM-crashes the vLLM engine
03 Exploitation and scoring
- Exploitation
- Not assessed. CISA has not published an exploitation assessment for this record, and it is not in the KEV catalog. That is an absence of assessment, not a finding of no risk.
- SSVC decision
- Not assessed by CISA
- EPSS probability
- 0.30% probability of exploitation activity in the next 30 days.
- EPSS percentile
- Equal to or higher than the score of 23.1% of scored records. A rank, not a probability.
- NVD base score
- 6.5 (CVSS v3.x), scored by the NVD independently of the CNA.
- NVD CVSS vector
- CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H
- NVD analysis status
- Analyzed. NVD last modified this record on 2026-10-28.
- Red Hat severity
- No Red Hat rating. Red Hat rates the CVEs that affect its products.
Exploit-prediction scores from FIRST. Independent scoring and analysis status from the NVD. Vendor severity from Red Hat. Exploitation assessment and catalog membership from CISA.
04 Metadata
- Published
- 2026-10-26 16:56Z2 DAYS AGO
- Last updated
- 2026-10-26 16:56Z2 DAYS AGO
- Reserved
- 2026-10-23
- Assigning CNA
- GitHub_M
- Record state
- PUBLISHED
- Severity
- Medium (CVSS 6.5)
- CNA CVSS vector
- CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H
- Weaknesses
- CWE-770Allocation of Resources Without Limits or ThrottlingBase
- Data version
- 5.2
- Document digest
- 02f20a2e42f6816b26a412f6925224e64755e3139694dfbf16f4bc1dc755d632
05 Affected products
VendorProductVersionsPlatforms
06 References
- github.comhttps://github.com/vllm-project/vllm/security/advisories/GHSA-8397-8qf6-f2gm
- github.comhttps://github.com/vllm-project/vllm/pull/60614
- github.comhttps://github.com/vllm-project/vllm/commit/3d52578e13454de0bb8352cac988347cdcc89928
- github.comhttps://github.com/vllm-project/vllm/releases/tag/v0.32.0