PyPI: sglang
CVE-2026-3059
Safety vulnerability ID: SFTY-20260312-16031
Safety legacy ID: pyup.io-89031
Affected versions of the sglang package are vulnerable to Deserialization of Untrusted Data due to unsafe deserialization of unauthenticated ZMQ broker messages with pickle.loads(). In python/sglang/multimodal_gen/runtime/scheduler_client.py, the multimodal generation broker receives attacker-controlled data with payload = await socket.recv() and immediately deserializes it with request_batch = pickle.loads(payload) without authentication or validation, which allows arbitrary Python objects to be reconstructed from untrusted input.
Overview
SGLang's multimodal generation module is vulnerable to unauthenticated remote code execution through the ZMQ broker
Advisory
Affected versions of the sglang package are vulnerable to Deserialization of Untrusted Data due to unsafe deserialization of unauthenticated ZMQ broker messages with pickle.loads(). In python/sglang/multimodal_gen/runtime/scheduler_client.py, the multimodal generation broker receives attacker-controlled data with payload = await socket.recv() and immediately deserializes it with request_batch = pickle.loads(payload) without authentication or validation, which allows arbitrary Python objects to be reconstructed from untrusted input.
How to Fix
Mitigation and Workarounds
---
Vulnerable Functions
Functions linked to known vulnerabilities.
References
- https://getsafety.com/vulnerabilities/SFTY-20260312-16031/CVE-2026-3059
- https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2026-3059
- https://github.com/advisories/GHSA-rgq9-fqf5-fv58
- https://github.com/sgl-project/sglang/security/advisories/GHSA-3cp7-c6q2-94xr
- https://nvd.nist.gov/vuln/detail/CVE-2026-3059
- https://github.com/sgl-project/sglang/blob/main/python/sglang/multimodal_gen/runtime/scheduler_client.py
- https://orca.security/resources/blog/sglang-llm-framework-rce-vulnerabilities
- https://github.com/advisories/GHSA-rgq9-fqf5-fv58
Verified by Safety
Our Cybersecurity Intelligence Team reviewed this vulnerability. We combine public data with our own research to find issues not yet reported to public sources.
Learn more