Vulnerability · Published Oct 16, 2026 · Assigned by VulnCheck
CVE-2026-111392SGLang through 0.5.21 KV Cache Poisoning via mm_hashes in /generate
Medium 4.8
In plain English
A medium-severity insufficient verification of data authenticity in sglang from sgl-project. Proof-of-concept exploit code is public. The chance it is exploited in the next 30 days is estimated at 0.2% (higher than 11% of all known vulnerabilities).
Severity (CVSS 3.1)4.8Medium
Exploit odds (30 days)0.2%top 89.3% of all CVEs
Exploited in the wildNo reportsCISA SSVC: poc
Attack toolingNoneno public tooling known
Official 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.
Affected products (as reported)
| Vendor | Product | Versions |
|---|---|---|
| sgl-project | sglang | 0.5.13 – through 0.5.21 |
Weakness: CWE-345 · Vector: CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:L/I:L/A:N
References
- https://github.com/sgl-project/sglang/issues/44641 issue-tracking
- https://github.com/sgl-project/sglang/blob/v0.5.21/python/sglang/srt/managers/tokenizer_manager.py#L1183-L1215 technical-description
- https://github.com/sgl-project/sglang/blob/v0.5.21/python/sglang/srt/managers/io_struct.py#L205-L215 technical-description
- https://github.com/sgl-project/sglang/pull/44653 patch, issue-tracking
- VulnCheck Advisory: SGLang through 0.5.21 KV Cache Poisoning via mm_hashes in /generate third-party-advisory