Windows Cloud Files CVE-2026-58536

Exploitation requires the attacker to already have local, authenticated access with at least low privileges (CVSS PR:L) and the ability to execute code on the target host - this is a post-compromise privilege escalation, not a remote entry point. The vulnerable component is the Cloud Files Mini Filter Driver (cldflt.sys), so the Cloud Files / Files On-Demand subsystem must be present and reachable, which it is by default on the listed Windows 10/11 and Windows Server 2019-2025 builds. Limiting factors: no network vector (AV:L) means the attacker cannot reach it remotely, and success depends on winning a use-after-free timing/heap-grooming condition in the kernel; however, no user interaction is required (UI:N) and attack complexity is low (AC:L).

The CVSS 3.1 base score is 7.8 (High) with vector AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H - local attack vector, low complexity, low privileges required, no user interaction, unchanged scope, and high impact to confidentiality, integrity, and availability. This is a realistic escalation primitive: any already-present low-privileged account (including a compromised standard user or malware post-initial-access) can attempt exploitation without user interaction, and success yields full SYSTEM control. Mitigating the urgency, exploitation requires local code execution first (no remote/network reach), no EPSS score was provided in the input, there is no public exploit identified, and it is not in CISA KEV - signals that suggest it is a valuable second-stage privilege escalation rather than an imminent, wormable mass-exploitation threat. The tags list 'Denial Of Service' as an additional possible outcome, consistent with the A:H availability impact of a kernel use-after-free.

An attacker who has already gained a foothold as a standard (low-privileged) user - via phishing, a compromised service account, or a malicious insider - runs a crafted local program that manipulates cloud placeholder file operations to force the cldflt.sys driver to dereference a freed kernel object. By grooming the kernel pool to reoccupy the freed allocation, the attacker corrupts kernel memory to gain arbitrary code execution as SYSTEM. No user interaction is required and attack complexity is low, though no public proof-of-concept exploit has been identified at time of analysis.