GLPI CVE-2026-47679

Requires a valid authenticated GLPI session at any privilege level (PR:L) - the flaw is in the profile-picture update flow available to ordinary self-service users, so no admin rights are needed. The affected feature (updating one's profile picture) is present in default installations of GLPI 10.0.0 through pre-10.0.26 and pre-11.0.8, so no non-default configuration is required. Exploitation is limited only by the need to log in first and by which files the web/application service account has permission to delete on the host filesystem.

This is a genuine, actionable vulnerability rather than a paper-high CVSS. The CVSS 4.0 vector (8.5) shows network access, low complexity, no user interaction, and only low privileges required (PR:L) - meaning ANY authenticated GLPI user (the lowest self-service account) can trigger it against the default configuration. The impact profile is deliberately integrity/availability-focused: VC:N (no data read), but VI:H/VA:H plus SI:H/SA:H, consistent with arbitrary file deletion that can corrupt application config/data and take the instance offline. No EPSS score, no CISA KEV listing, and no public exploit code were provided, so this is 'no public exploit identified at time of analysis' - the risk is driven by ease of exploitation and low auth bar, not by observed in-the-wild activity. The main mitigating signal is that exploitation requires a valid login (PR:L), so it is not internet-unauthenticated. The subsequent-system impact metrics (SI:H/SA:H) in the vendor vector are somewhat aggressive since the affected asset IS the GLPI server itself (no true scope change to a separate system); this does not change the practical severity but is worth noting.

A low-privileged but authenticated GLPI user submits a crafted profile-picture update request whose file path traverses outside the intended avatar directory (e.g. pointing at a GLPI config file, database artifact, or plugin file). The insufficient path validation causes the server to delete that attacker-chosen file, corrupting the installation or causing denial of service. No public exploit code was provided, but the patch (adding a validation return false) makes the flawed flow easy to reconstruct.