# CVE-2026-111160: LightLLM denial of service via pd_kv_trans_params unpickling

Source: https://junglewise.ai/threats/cve-2026-111160-lightllm-through-1-2-0-contains-a-denial-of-service  
Publisher: Junglewise Threat Intelligence, https://junglewise.ai/threats  
Published: 2026-10-15. Updated: 2026-10-21  
License: CC BY 4.0 with a link back

CVE-2026-111160 · Severity: high · CVSS 7.5 · Published 2026-10-15

LightLLM is a Python-based inference framework for large language models. An unauthenticated attacker can crash the inference worker thread by sending a specially crafted request to the /generate API endpoint, causing all subsequent requests to hang indefinitely while the health check falsely reports the service as operational. This allows an attacker to disable the service with a single malicious request without any special credentials or internal network access.

Technologies: [ModelTC LightLLM](https://junglewise.ai/threats/technologies/lightllm). Vendors: [ModelTC](https://junglewise.ai/threats/vendors/modeltc).

## Quick answers

**Is CVE-2026-111160 exploited in the wild?** No. Junglewise Threat Intelligence has no report of CVE-2026-111160 being exploited in the wild.

**How severe is CVE-2026-111160?** High severity, with a CVSS score of 7.5, published 15 October 2026.

**Which products and versions are affected?** ModelTC LightLLM through 1.2.0.

**Is there a fix for CVE-2026-111160?** 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-2026-111160 attacked?** It is attacked over the network, in a software library.

**Am I affected by CVE-2026-111160?** Check whether you run ModelTC LightLLM 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 handling pipeline where client-controlled pd_kv_trans_params input is passed through SamplingParams initialization and later deserialized via pickle.loads in InferSamplingParams without validation. An attacker sends a non-empty string (e.g., "AAAAdeadbeefBBBB") as the pd_kv_trans_params field, which is copied into shared memory as UTF-32 encoded bytes that fail pickle deserialization, raising an UnpicklingError. The exception occurs in the inference worker thread, terminating it while the HTTP process and health endpoint remain responsive, causing subsequent requests to hang indefinitely.

## Affected products

- ModelTC LightLLM through 1.2.0

## Timeline

- 2026-10-15: disclosed

## References

- [https://github.com/ModelTC/LightLLM](https://github.com/ModelTC/LightLLM)
- [https://github.com/ModelTC/LightLLM/issues/1629](https://github.com/ModelTC/LightLLM/issues/1629)

## Related threats

- [LightLLM visual_only unauthenticated RCE via RPyC pickle deserialization](https://junglewise.ai/threats/cve-2026-103395-lightllm-through-1-2-0-visual-only-deployments-expose-an)
- [LightLLM authentication bypass in reinforcement learning control API](https://junglewise.ai/threats/cve-2026-103270-lightllm-through-1-2-0-mounts-reinforcement-learning-control)
- [LightLLM server-side request forgery in multimodal endpoints](https://junglewise.ai/threats/cve-2026-103243-lightllm-through-1-2-0-fails-to-validate-image-url-and-audio-url)
- [LightLLM memory exhaustion in NCCL control channel](https://junglewise.ai/threats/cve-2026-103042-lightllm-through-1-2-0-contains-a-memory-exhaustion)
- [LightLLM multimodal pickle deserialization in RPyC cache service](https://junglewise.ai/threats/cve-2026-103041-lightllm-through-1-2-0-multimodal-deployments-expose-an)