Microsoft COM for Windows EUVD-2026-73434

The attacker must hold a low-privileged authenticated local session on the affected Windows host (PR:L in CVSS); remote unauthenticated exploitation is not described. The attack complexity is rated High (AC:H), indicating a prerequisite of either a race-condition timing window or specific heap-layout conditions that are not reliably reproducible without deliberate setup. No specific non-default feature or configuration flag is cited as a precondition - COM is a default component present in all listed Windows versions. Limiting factors: requires local code execution as a standard user (not guest), requires winning a race or achieving heap-grooming success, and exploitation is constrained to the host's privilege boundary (S:U, no cross-process/network scope change).

The CVSS 7.0 score (AV:L/AC:H/PR:L/UI:N/S:U/C:H/I:H/A:H) accurately reflects a meaningful but non-trivial local privilege escalation: an attacker must already hold a low-privileged local account and must successfully exploit a race-condition or precise heap-grooming scenario (AC:H). The full C:H/I:H/A:H impact block is justified because successful exploitation yields SYSTEM-level access. No EPSS data was provided, KEV status is not confirmed, and no public PoC has been identified, which collectively suggest this is not under active widespread exploitation. The primary real-world risk profile is an insider threat or post-initial-compromise lateral/vertical movement scenario - an attacker who has established a foothold on a Windows host via phishing or another initial-access vector would chain this flaw to escape a restricted user context. The ubiquitous deployment of Windows across enterprise environments raises the aggregate exposure even if per-host exploitation complexity is high.

An attacker who has achieved a low-privileged shell on a Windows workstation - via phishing, a compromised application, or stolen credentials - triggers a race condition in the COM subsystem by rapidly allocating and releasing COM interface pointers while a concurrent operation accesses the freed memory. By timing or heap-grooming the reuse of the freed region with attacker-controlled data, execution is redirected to arbitrary code running as SYSTEM. No public proof-of-concept has been identified, indicating that reliable exploitation likely requires significant research or tooling effort.