PatchSiren cyber security CVE debrief
CVE-2026-111392 sgl-project CVE debrief
AI-assisted PatchSiren debrief based on the supplied source corpus. The CVE record was published on 2026-10-16T17:38:44.867Z and has not been modified since then. SGLang through 0.5.21 is vulnerable to KV Cache Poisoning via mm_hashes in /generate, allowing remote attackers to poison cached KV entries across requests. This could lead to disclosure of sensitive image content or corruption of victim responses. Defenders should assess exposure, especially in SGLang deployments using version 0.5.21 or earlier, and verify cache management configurations.
- Vendor
- sgl-project
- Product
- sglang
- CVSS
- MEDIUM 6.3
- CISA KEV
- Not listed in stored evidence
- Original CVE published
- 2026-10-16
- Original CVE updated
- 2026-10-16
- Advisory published
- 2026-10-16
- Advisory updated
- 2026-10-16
Who should care
Defenders responsible for SGLang deployments, especially those using version 0.5.21 or earlier, should assess exposure and verify cache management configurations. This includes operators, platform administrators, vulnerability management teams, and security teams who need to prioritize verifying exposure and implementing compensating controls.
Why it matters
CVE-2026-111392 allows remote attackers to poison cached KV entries in SGLang through 0.5.21, potentially disclosing sensitive image content or corrupting victim responses. Defenders should prioritize verifying exposure, especially in SGLang deployments using version 0.5.21 or earlier, and assess the feasibility of cache poisoning attacks. Additional verification is needed to confirm affected versions and potential impact.
- Potential disclosure of sensitive image content
- Corruption of victim responses
- Cache poisoning attacks may require verification of affected versions and configurations
- Defenders should prioritize verifying exposure and implementing compensating controls
Technical summary
SGLang through 0.5.21 trusts caller-supplied mm_hashes values as the multimodal prefix-cache key, allowing remote attackers to poison cached KV entries across requests. This could lead to disclosure of sensitive image content or corruption of victim responses. Defenders should assess exposure, especially in SGLang deployments using version 0.5.21 or earlier, and verify cache management configurations. Technical framing indicates that cache poisoning attacks may require verification of affected versions and configurations.
Defensive priority
Defenders should prioritize verifying exposure in SGLang deployments, especially those using version 0.5.21 or earlier, and assess the feasibility of cache poisoning attacks.
Recommended defensive actions
- Verify SGLang version and assess exposure to cache poisoning attacks
- Review and update cache management configurations
- Monitor for suspicious activity and implement compensating controls
- Confirm whether affected product deployments exist in managed environments and assign an owner for follow-up
- Review the supplied official advisory or CVE record to validate affected scope, severity, and vendor guidance
- Plan vendor-supported updates or mitigations through normal change control where exposure is confirmed
- Check relevant monitoring, detection, and logs for exposed assets that need extra review
Evidence notes
The CVE record and source item provide details on the vulnerability, but additional verification is needed to confirm affected versions and potential impact. Source grounding indicates that SGLang through 0.5.21 trusts caller-supplied mm_hashes values, and defenders should verify cache management configurations. Evidence limits suggest that further review of cache poisoning attacks and affected scope is necessary.
Sources and references
Verified primary and authoritative sources
-
CVE-2026-111392 CVE Program record
Publisher, destination, and source semantics verified
URL: https://www.cve.org/CVERecord?id=CVE-2026-111392
CVE Program - Official CVE Program record with source-provided CVE metadata.
-
CVE-2026-111392 NVD vulnerability detail
Publisher, destination, and source semantics verified
URL: https://nvd.nist.gov/vuln/detail/CVE-2026-111392
NIST National Vulnerability Database - Official NIST NVD detail page and source-specific vulnerability assessment.
Supplemental references
-
SGLang through 0.5.21 KV Cache Poisoning via mm_hashes in /generate
Unverified legacy reference
URL: https://raw.githubusercontent.com/CVEProject/cvelistV5/main/cves/2026/111xxx/CVE-2026-111392.json
cve_program_cvelist_v5
-
Source reference
Unverified legacy reference
URL: https://github.com/sgl-project/sglang/issues/44641
Supplemental source - issue-tracking
-
Source reference
Unverified legacy reference
URL: https://github.com/sgl-project/sglang/blob/v0.5.21/python/sglang/srt/managers/tokenizer_manager.py
Supplemental source - technical-description
-
Source reference
Unverified legacy reference
URL: https://github.com/sgl-project/sglang/blob/v0.5.21/python/sglang/srt/managers/io_struct.py
Supplemental source - technical-description
-
Source reference
Unverified legacy reference
URL: https://github.com/sgl-project/sglang/pull/44653
Supplemental source - patch, issue-tracking
-
Source reference
Unverified legacy reference
URL: https://www.vulncheck.com/advisories/sglang-through-0.5.21-kv-cache-poisoning-via-mm-hashes-in-generate
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.