CVE-2026-111392
4.8 MEDIUMpublic exploit availablePublished 2026-10-16 · Updated 2026-10-16
AI analysis for this CVE has not been generated yet. Raw NVD data is shown below.
NVD description
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. Attackers can send a different image under a known or predicted hash via /generate to disclose another request's image content or corrupt victims' responses.
CVSS vector
CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:L/I:L/A:N
Weaknesses
CWE-345
Public exploit & PoC references
- https://github.com/sgl-project/sglang/blob/v0.5.21/python/sglang/srt/managers/io_struct.py#L205-L215
- https://github.com/sgl-project/sglang/blob/v0.5.21/python/sglang/srt/managers/tokenizer_manager.py#L1183-L1215
- https://github.com/sgl-project/sglang/issues/44641
- https://github.com/sgl-project/sglang/pull/44653
- https://github.com/sgl-project/sglang/issues/44641
All references
- https://github.com/sgl-project/sglang/blob/v0.5.21/python/sglang/srt/managers/io_struct.py#L205-L215
- https://github.com/sgl-project/sglang/blob/v0.5.21/python/sglang/srt/managers/tokenizer_manager.py#L1183-L1215
- https://github.com/sgl-project/sglang/issues/44641
- https://github.com/sgl-project/sglang/pull/44653
- https://www.vulncheck.com/advisories/sglang-through-0.5.21-kv-cache-poisoning-via-mm-hashes-in-generate
- https://github.com/sgl-project/sglang/issues/44641
Source data: NVD (nvd.nist.gov), public domain. Exploit-DB.ai adds local AI analysis for defensive use only.
Related CVEs
- UNSCOREDCVE-2025-12999PoC
- HIGHCVE-2026-100872PoC
- MEDIUMCVE-2026-101278
- MEDIUMCVE-2026-102140PoC
- UNSCOREDCVE-2026-102250
- HIGHCVE-2026-102267PoC
- MEDIUMCVE-2026-102275PoC
- HIGHCVE-2026-102677PoC
Related by shared AI tags and CWE weakness class. Browse the full archive.