# CVE-2026-90919: LightLLM Config Server unsafe pickle deserialization in /visual_register

Source: https://junglewise.ai/threats/cve-2026-90919-lightllm-config-server-unsafe-pickle-deserialization-in-visual  
Publisher: Junglewise Threat Intelligence, https://junglewise.ai/threats  
Published: 2026-09-14. Updated: 2026-09-26  
License: CC BY 4.0 with a link back

CVE-2026-90919 · Severity: critical · CVSS 9.8 · Published 2026-09-14

LightLLM's Config Server is a control-plane service that coordinates deployment and configuration of machine-learning inference workloads. The /visual_register WebSocket endpoint accepts unauthenticated connections and deserializes untrusted data using Python's pickle module, allowing remote attackers to execute arbitrary code with Config Server process privileges. This compromises the entire cluster orchestration layer.

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

## Quick answers

**Is CVE-2026-90919 exploited in the wild?** No. Junglewise Threat Intelligence has no report of CVE-2026-90919 being exploited in the wild.

**How severe is CVE-2026-90919?** Critical severity, with a CVSS score of 9.8, published 14 September 2026.

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

**Is there a fix for CVE-2026-90919?** 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-2026-90919 attacked?** It is attacked over the network, in a software library.

**Am I affected by CVE-2026-90919?** 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 is an unsafe deserialization flaw (CWE-502) in the /visual_register WebSocket endpoint of LightLLM's Config Server (lightllm/server/config_server/api_http.py, line 83). The handler accepts WebSocket connections without authentication, immediately calls pickle.loads() on the first client-supplied binary frame, and stores the deserialized object in a registry. Because pickle can execute arbitrary Python code during reconstruction via __reduce__ methods and other mechanisms, an attacker with network access to the Config Server port can send a malicious serialized object to achieve remote code execution with the service process privileges. No authentication or validation occurs before deserialization, and the VIT_Obj type annotation provides no runtime security enforcement. The issue affects LightLLM through version 1.2.0.

## Affected products

- ModelTC LightLLM through 1.2.0

## Timeline

- 2026-09-14: disclosed
- 2026-09-14: advisory

## References

- [https://github.com/ModelTC/LightLLM/blob/v1.2.0/lightllm/server/config_server/api_http.py](https://github.com/ModelTC/LightLLM/blob/v1.2.0/lightllm/server/config_server/api_http.py)
- [https://github.com/ModelTC/LightLLM/issues/1563](https://github.com/ModelTC/LightLLM/issues/1563)

## Related threats

- [LightLLM denial of service via pd_kv_trans_params unpickling](https://junglewise.ai/threats/cve-2026-111160-lightllm-through-1-2-0-contains-a-denial-of-service)
- [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)