FortiDLP Agent CVE-2025-53950

Exploitation requires the FortiDLP Agent's Outlookproxy plugin to be installed and active on a MacOS or Windows endpoint, AND the attacker must already be an authenticated administrator with local access to that host (AV:L/PR:H). Affected versions are 11.5.1, 11.4.2-11.4.6, 11.3.2-11.3.4, 11.2.0-11.2.3, 11.1.1-11.1.2, 11.0.1, 10.5.1, 10.4.0, and 10.3.1. The limiting factors are substantial: no remote vector, no unauthenticated path, and the requirement for high (administrator) privileges mean a non-privileged or remote attacker cannot exploit this; realistic risk is confined to insider misuse or an attacker who has already achieved local administrative compromise.

This is a high-CVSS-component-but-low-real-risk issue in practice. The CVSS 3.1 vector (AV:L/AC:L/PR:H/UI:N/S:C/C:H/I:N/A:N = 6.0 Medium) is the dominant honest signal: exploitation requires LOCAL access and HIGH privileges - specifically an authenticated administrator - which already presumes a largely trusted, elevated actor. The scope-change (S:C) and C:H elevate the base score because the administrator can reach another principal's (the current user's) email data, but no integrity or availability impact exists. Supporting signals reinforce low urgency: EPSS is 0.18% (7th percentile), there is no CISA KEV listing, and no public exploit code was identified. These signals are consistent with each other - low EPSS, no KEV, no POC, local vector, and high-privilege precondition all point the same direction, so there is no conflict to resolve. The only data gap is that the Fortinet PSIRT advisory (FG-IR-25-639) content was not provided for independent confirmation of fixed versions or exploitation claims. Bottom line: this is a privacy/confidentiality disclosure exploitable only by an already-privileged local admin - a defense-in-depth / insider-risk concern, not an urgent remote threat.

An authenticated administrator with local access to a MacOS or Windows host running the FortiDLP Agent abuses the Outlookproxy plugin to read the current logged-in user's email information, which they would not normally be entitled to access. No public exploit code is known, and exploitation is constrained by the requirement for pre-existing high privileges and local access (AV:L/PR:H), making this primarily an insider or post-compromise privacy-violation scenario rather than a remotely weaponizable flaw.