CVE-2026-111391 PoC?

SGLang 0.5.10 through 0.5.22 contains a code injection vulnerability in _evaluate_filter() that allows remote attackers to execute Python code by escaping the eval() sandbox through re.search globals. Attackers can send a crafted filter expression to the unauthenticated /dumper/configure endpoint when DUMPER_SERVER_PORT is set, running code in the scheduler process on the next forward pass.

This vulnerability allows an attacker from the Internet to exploit an Improper Neutralization of Directives in Dynamically Evaluated Code ('Eval Injection') (CWE-95) weakness. A successful attack leads to a high loss of confidentiality, a high loss of integrity and a high loss of availability.

AV:N/AC:L/AT:P/PR:N/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N

Confidentiality Impact to the Vulnerable System

Integrity Impact to the Vulnerable System

Availability Impact to the Vulnerable System

Confidentiality Impact to the Subsequent System

Integrity Impact to the Subsequent System

Availability Impact to the Subsequent System

Temporal score updated on 18 Oct 2026, 03:09

Source: PatchCVE CVSS-BT — score adjusted with the official CVSS 4.0 algorithm (MacroVector) from the NVD base vector and the estimated Exploit Maturity (CISA KEV, exploit-db).

SSVC (CISA): Exploitation = PoC — public proof of concept observed

Temporal score updated on 18 Oct 2026, 03:09

SSVC (CISA): Exploitation = PoC — public proof of concept observed

CNA affected data: only versions up to 0.5.22 are affected — a fixed version exists above it