GL-iNet GL-MT3000 EUVD-2026-52397

The vulnerable attack surface is the /cgi-bin/glc CGI endpoint, specifically its plugins.set_config handler within plugins.so. Per the reported CVSS 4.0 vector (PR:N/UI:N), no authentication or user interaction is required; however, this claim has not been independently verified against firmware analysis and conflicts with the typical authentication model of GL-iNet CGI management endpoints - if session authentication is in fact enforced, an attacker would need valid admin credentials, substantially limiting exposure. The primary real-world limiting factor is network reachability: by default on consumer deployments, the management interface is not exposed to the WAN, restricting exploitation to local network attackers or those who have already achieved LAN access. Remote exploitation is viable only when WAN-side management is explicitly enabled. The vulnerability is firmware-version bounded to 4.4.5 and earlier on the GL-MT3000 platform specifically.

The CVSS 4.0 base score of 8.9 with AV:N/AC:L/AT:N/PR:N/UI:N and full H/H/H impact signals a high-severity, low-friction remote exploitation path. The E:P modifier confirms proof-of-concept code is publicly available on GitHub, materially lowering the skill threshold for exploitation. The vulnerability is not currently listed in CISA KEV, meaning active widespread exploitation has not been formally confirmed; however, consumer and prosumer routers with public PoC and unpatched injection flaws historically attract botnet operators within days to weeks of disclosure. EPSS data was not provided, limiting probabilistic exploitation forecasting. A critical caveat: the PR:N assignment should be independently verified - GL-iNet CGI management endpoints typically enforce session authentication, and if this endpoint silently enforces auth, real-world unauthenticated exploitability would be substantially overstated by the reported vector. If PR:N is accurate, this is a genuine priority for any network exposing the device management interface to untrusted segments.

An attacker with network access to the GL-MT3000 management interface - either on the LAN or via the internet if remote management is enabled - sends a crafted HTTP request to /cgi-bin/glc invoking plugins.set_config with a malicious parameter containing an injection payload. The plugins.so native library processes the input without sanitization, causing the injected content to be interpreted as an OS command or argument within the router's execution context. Because a public proof-of-concept targeting firmware 4.4.5 is available on GitHub, this attack can be reproduced with minimal effort by a low-skilled attacker.