GLPI CVE-2025-59935

Exploitation targets GLPI's inventory endpoint (the agent/native inventory ingestion interface), which can accept data from unauthenticated sources - this is the specific feature required, so environments that expose native inventory intake to untrusted networks are most at risk. The attack is a two-stage stored XSS: the attacker can inject without credentials (PR:N), but the payload only fires when an authenticated GLPI user subsequently renders the poisoned inventory record in the web UI (UI:R), which is the key limiting factor. Risk is reduced where the inventory endpoint is network-restricted, requires agent authentication/tokens, or is not exposed to untrusted hosts; affected versions are GLPI 10.0.0 through 10.0.20, fixed in 10.0.21.

This is a moderate-priority vulnerability rather than an emergency, but the unauthenticated storage vector raises its practical concern above the 6.5 CVSS base. The CVSS vector (AV:N/AC:L/PR:N/UI:R) shows a network-reachable, low-complexity flaw that requires no attacker privileges but does require victim interaction (an authenticated user viewing the injected inventory data). EPSS is low at 0.29% (20th percentile), indicating no statistical signal of imminent mass exploitation, and the issue is NOT on CISA KEV, so there is no evidence of active exploitation. No POC was provided in the intelligence set. The one notable tension is impact scoring: the supplied vector rates confidentiality as C:H with S:U, which is unusually high for a stored XSS where the script typically executes in a victim's browser session and would more conventionally map to a scope change with lower, split C/I impact - this discrepancy should be verified against the GitHub advisory (GHSA-j8vv-9f8m-r7jx). Overall: a real and patchable issue worth prioritizing for internet-exposed GLPI instances, but not a drop-everything KEV-class threat.

An unauthenticated attacker submits a crafted inventory payload to GLPI's inventory endpoint containing a stored XSS script. When a logged-in GLPI operator or administrator later opens the affected asset/inventory record in the web console, the script executes in their authenticated session, enabling session hijacking or actions performed as that user. No public exploit code was identified at time of analysis, and successful execution depends on an authenticated user viewing the malicious entry (UI:R).