SGLang 0.5.22 及更早版本在使用 Mooncake 解耦后端时,其 函数中存在一个不受控的内存分配漏洞。该漏洞允许未认证的邻近网络攻击者耗尽 Prefill 内存。攻击者可通过 接口获取 rank 地址,并针对新的 room ID 发送包含巨大 值的 消息,导致调度器因内存不足(OOM)而被系统终止。
Although we use advanced large model technology, its output may still contain inaccurate or outdated information.Shenlong tries to ensure data accuracy, but please verify and judge based on the actual situation.
| Vendor | Product | Affected Versions | CPE | Subscribe |
|---|---|---|---|---|
| sgl-project | sglang | 0 ~ 0.5.22 | - |
|
| # | POC Description | Source Link | Shenlong Link |
|---|
No public POC found.
Login to generate AI POC| CVE-2026-112082 | 7.5 HIGH | sgl-project SGLang Expert-Distribution Recording Endpoint expert_distribution.py Scheduler |
| CVE-2026-112267 | 6.5 MEDIUM | SGLang through 0.5.22 SSRF via bootstrap_host and bootstrap_port on Decode Worker |
| CVE-2026-112269 | 6.5 MEDIUM | SGLang through 0.5.22 Memory Exhaustion via Unauthenticated Mooncake Decode Registrations |
| CVE-2026-112268 | 6.3 MEDIUM | sglang-router through 0.3.2 Authorization Bypass via Mesh HA Management Routes |
| CVE-2026-112266 | 5.9 MEDIUM | SGLang through 0.5.22 Cross-Request Abort via rid Prefix Matching |
| CVE-2026-112083 | 5.3 MEDIUM | sgl-project SGLang Weight Exporter weight_exporter.py WeightExporter.send_weights_to_remot |
No comments yet