PatchSiren cyber security CVE debrief
CVE-2026-109227 sgl-project CVE debrief
A memory exhaustion denial-of-service vulnerability exists in SGLang through version 0.5.21. The issue arises from unbounded memory allocation in the preprocess_video function of qwen_vl.py, specifically when decoding frames at full source resolution before applying max_pixels limits. This allows remote attackers to submit high-resolution videos to /v1/chat/completions on Qwen-VL deployments. Consequently, concurrent requests can exhaust worker memory, leading to denial of service.
- Vendor
- sgl-project
- Product
- sglang
- CVSS
- HIGH 8.7
- CISA KEV
- Not listed in stored evidence
- Original CVE published
- 2026-10-12
- Original CVE updated
- 2026-10-12
- Advisory published
- 2026-10-12
- Advisory updated
- 2026-10-12
Who should care
Defenders responsible for SGLang deployments, particularly those using Qwen-VL, should assess their exposure to untrusted video inputs and verify the version of SGLang in use. This vulnerability may impact systems that process video inputs from untrusted sources.
Why it matters
CVE-2026-109227 is a memory exhaustion denial-of-service vulnerability in SGLang through version 0.5.21. It affects Qwen-VL deployments and allows remote attackers to cause denial of service by submitting high-resolution videos. Defenders should verify exposure, assess upgrade feasibility, and implement input validation and rate limiting.
- Denial of service through memory exhaustion
- Potential for system crashes or instability
- Need for input validation and rate limiting
- Verification of SGLang version and configuration
Technical summary
The vulnerability exists in the preprocess_video function of qwen_vl.py in SGLang through version 0.5.21. It allows for unbounded memory allocation by decoding frames at full source resolution before applying max_pixels limits. This can be exploited by remote attackers submitting high-resolution videos to /v1/chat/completions on Qwen-VL deployments, leading to denial of service through memory exhaustion.
Defensive priority
Defenders should prioritize verifying exposure of SGLang deployments to untrusted video inputs and assess the feasibility of upgrading to a fixed version.
Recommended defensive actions
- Verify SGLang deployment versions and configurations to identify potential exposure
- Assess the feasibility of upgrading to a fixed version of SGLang
- Implement input validation and limits for video processing to prevent memory exhaustion
- Monitor for concurrent requests and implement rate limiting if necessary
- Review compensating controls for exposed systems while remediation is scheduled and verified
- Check relevant monitoring, detection, and logs for exposed assets that need extra review
- Track exceptions, retest remediated assets, and close the item only after evidence is documented
Evidence notes
The CVE record and source item provide details on the vulnerability, including its description, affected versions, and references to technical descriptions and issue tracking. Defenders should verify the version of SGLang in use, assess exposure to untrusted video inputs, and implement input validation and rate limiting. The preprocess_video function in qwen_vl.py decodes frames at full source resolution before applying max_pixels limits, allowing for unbounded memory allocation. This can be exploited by remote attackers submitting高清.
Sources and references
Verified primary and authoritative sources
-
CVE-2026-109227 CVE Program record
Publisher, destination, and source semantics verified
URL: https://www.cve.org/CVERecord?id=CVE-2026-109227
CVE Program - Official CVE Program record with source-provided CVE metadata.
-
CVE-2026-109227 NVD vulnerability detail
Publisher, destination, and source semantics verified
URL: https://nvd.nist.gov/vuln/detail/CVE-2026-109227
NIST National Vulnerability Database - Official NIST NVD detail page and source-specific vulnerability assessment.
Supplemental references
-
SGLang through 0.5.21 Memory Exhaustion DoS via Qwen-VL Video Preprocessing
Unverified legacy reference
URL: https://raw.githubusercontent.com/CVEProject/cvelistV5/main/cves/2026/109xxx/CVE-2026-109227.json
cve_program_cvelist_v5
-
Source reference
Unverified legacy reference
URL: https://github.com/sgl-project/sglang/issues/43710
Supplemental source - issue-tracking
-
Source reference
Unverified legacy reference
URL: https://github.com/sgl-project/sglang/blob/v0.5.21/python/sglang/srt/multimodal/processors/qwen_vl.py
Supplemental source - technical-description
-
Source reference
Unverified legacy reference
URL: https://github.com/sgl-project/sglang
Supplemental source - product
-
Source reference
Unverified legacy reference
URL: https://www.vulncheck.com/advisories/sglang-through-0.5.21-memory-exhaustion-dos-via-qwen-vl-video-preprocessing
Supplemental source - third-party-advisory
Methodology and review provenance
AI-assisted synthesis based on stored public vulnerability evidence. System validation, approval state, and publication status do not by themselves establish human review of this revision. PatchSiren helps prioritize defensive review and does not prove exposure or remediation on any system.