Windows Audio Service EUVD-2026-73013

Exploitation requires an authenticated local user account (PR:L); the attack vector is strictly local (AV:L), making remote unauthenticated exploitation impossible. The high attack complexity (AC:H) reflects a race condition: the attacker must win a timing-dependent window between object deallocation and reuse in the Audio Service's multi-threaded execution, requiring repeated attempts or careful synchronization primitives. Windows Audio Service (AudioSrv) must be running - it is enabled and started by default on all Windows desktop editions but may be disabled on hardened Windows Server deployments that have no audio requirement, in which case the attack surface is absent. No special configuration or non-default feature is required on standard workstation builds. Limiting factors include: the local-only attack vector restricting exploitation to already-present users, and the AC:H race condition reducing reliability and preventing trivially automatable mass exploitation.

The CVSS 7.0 score with AV:L/AC:H/PR:L/UI:N/S:U/C:H/I:H/A:H accurately reflects this as a local privilege escalation with meaningful exploitation difficulty - the AC:H metric captures the race-condition requirement, which substantially reduces automated and opportunistic exploitability. EPSS data was not supplied; however, Windows LPE vulnerabilities in core system services are historically attractive as post-exploitation primitives for ransomware operators and nation-state actors who already hold an initial foothold. No public exploit code has been observed and CISA has not added this to the Known Exploited Vulnerabilities catalog, suggesting exploitation remains private or theoretical at this stage. The S:U (scope unchanged) and PR:L metrics bound the attacker to an already-authenticated, already-present user on the system, meaning this flaw is a second-stage capability rather than an initial access vector. Defenders should prioritize patch deployment on workstations and endpoints with untrusted or shared user access, as those are the environments where low-privileged attackers are most likely to already exist.

An attacker with a standard local user account - obtained via phishing, credential theft, or lateral movement - identifies the race condition window in Windows Audio Service's memory allocation path. By racing concurrent API calls against the Audio Service to trigger the use-after-free condition, the attacker controls the freed memory region to redirect code execution within the privileged service context, yielding SYSTEM-level access. No public proof-of-concept exists, but the race condition requirement means the attacker may need to run the attempt repeatedly to succeed, which may generate detectable anomalous Audio Service API activity.