Windows ProjFS EUVD-2026-35533

Exploitation requires the attacker to already hold valid low-privileged interactive or programmatic access on the target Windows host (CVSS PR:L, AV:L) and requires the Windows Projected File System optional feature to be enabled so that prjflt.sys is loaded and reachable - ProjFS is not enabled by default on all SKUs and is most commonly present on developer workstations and build servers running VFS for Git/Scalar. No user interaction is needed (UI:N), and attack complexity is low (AC:L), but remote network-only attackers without a local account cannot exploit this directly, and systems where the ProjFS feature is uninstalled are not exposed.

The CVSS 3.1 base score of 7.8 with vector AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H indicates a local, low-complexity attack requiring an already-authenticated low-privileged user, with no user interaction, and full CIA impact within the same scope - a textbook local elevation of privilege profile. No EPSS score, KEV listing, or POC reference was supplied in the input, so the realistic risk at this moment is moderate: the bug is meaningful for any multi-user, kiosk, terminal-server, or post-initial-access scenario, but there is no evidence of active exploitation or weaponization. Note a tension in the public signals - the tag labels this a 'Buffer Overflow' while the CWE-126 mapping is specifically a buffer over-read, which is more typical of an info-leak primitive than direct code execution; the C:H/I:H/A:H rating suggests Microsoft assesses it can be driven beyond a pure read, but the exact exploitation primitive should be verified against MSRC's technical write-up when published.

An attacker first obtains low-privileged code execution on a Windows host - for example, via a phished user account, a compromised service account, or a foothold from another vulnerability - then runs a small native binary that opens a handle to the ProjFS driver and issues crafted IOCTL or callback inputs that trigger the over-read. The leaked kernel memory is used to defeat KASLR and/or pivot into an elevation primitive, ultimately yielding SYSTEM-level execution that enables credential theft and lateral movement. No public exploit identified at time of analysis, so this scenario is the expected post-disclosure trajectory rather than an observed campaign.