# CVE-2026-112266: SGLang denial of service via request ID prefix matching

Source: https://junglewise.ai/threats/cve-2026-112266-sglang-through-0-5-22-contains-a-denial-of-service-vulnerability  
Publisher: Junglewise Threat Intelligence, https://junglewise.ai/threats  
Published: 2026-10-19. Updated: 2026-10-19  
License: CC BY 4.0 with a link back

CVE-2026-112266 · Severity: medium · CVSS 5.9 · Published 2026-10-19

SGLang is a language model serving framework that handles API requests for text generation. A vulnerability in versions through 0.5.22 allows an attacker to disrupt the service for other users by submitting crafted requests with short identifiers, causing all in-flight generation requests to be aborted. This results in service denial and loss of work for legitimate users.

Technologies: [SGLang](https://junglewise.ai/threats/technologies/sglang-sglang). Vendors: [SGLang](https://junglewise.ai/threats/vendors/sglang).

## Quick answers

**Is CVE-2026-112266 exploited in the wild?** No. Junglewise Threat Intelligence has no report of CVE-2026-112266 being exploited in the wild.

**How severe is CVE-2026-112266?** Medium severity, with a CVSS score of 5.9, published 19 October 2026.

**Which products and versions are affected?** SGLang through 0.5.22.

**Is there a fix for CVE-2026-112266?** No fixed version is recorded yet. Check the vendor's advisory in the references below. If you run it, Junglewise drafts a mitigation that fits your setup and applies it once you approve, with senior security engineers behind it.

**How is CVE-2026-112266 attacked?** It is attacked over the network, in a software library.

**Am I affected by CVE-2026-112266?** Check whether you run SGLang at an affected version. Junglewise does that check for you around the clock: connect your code, cloud and domains, and it matches new vulnerabilities like this one against what you actually run and tells you when one reaches you.

## Technical details

The vulnerability exists in the request abort logic within SGLang's scheduler, which uses prefix-matching on client-supplied request IDs (rid) instead of exact matching. An attacker can submit a streaming /generate request with a single hex character rid and immediately disconnect, triggering the prefix-match logic to terminate all other clients' requests whose IDs share that prefix. This requires network access to the API endpoint but no authentication or special privileges.

## Affected products

- SGLang SGLang through 0.5.22

## Timeline

- 2026-10-19: disclosed

## References

- [https://github.com/sgl-project/sglang/blob/v0.5.22/python/sglang/srt/managers/scheduler.py](https://github.com/sgl-project/sglang/blob/v0.5.22/python/sglang/srt/managers/scheduler.py)
- [https://github.com/sgl-project/sglang/blob/v0.5.22/python/sglang/srt/managers/tokenizer_manager.py](https://github.com/sgl-project/sglang/blob/v0.5.22/python/sglang/srt/managers/tokenizer_manager.py)

## Related threats

- [SGLang unbounded memory allocation in video preprocessing](https://junglewise.ai/threats/cve-2026-109227-sglang-through-0-5-21-contains-an-unbounded-memory-allocation)
- [SGLang unbounded memory allocation in audio transcription endpoint](https://junglewise.ai/threats/cve-2026-109225-sglang-through-0-5-21-contains-an-unbounded-memory-allocation)
- [SGLang Decode worker DoS in ZeroMQ control message handling](https://junglewise.ai/threats/cve-2026-94570-sglang-contains-a-dos-vulnerability-caused-by-missing-input)
- [SGLang unauthenticated arbitrary code execution in multimodal orchestrator](https://junglewise.ai/threats/cve-2026-93088-sglang-s-multimodal-generation-runtime-is-vulnerable-to)