GBHackers

Apple Microphone Vulnerability Allows Hackers to Secretly Record Audio Without User Permission


A flaw in Apple’s audio media daemon let sandboxed apps record microphone audio without permission and, on an earlier build, without the orange privacy indicator.

Meta Red Team X researcher Nik Tsytsarkin disclosed the bug as CVE-2026-65339, also crediting Mustafa Calap. Apple’s advisory describes it only as an Audio issue where “an app may be able to leak sensitive user information”.

Apple Microphone Vulnerability

The flaw sits in audiomxd, which runs the server side of AURemoteIO, the audio unit used for low-latency recording. Apple’s Transparency, Consent, and Control (TCC) framework normally decides whether a process may use the microphone, and that decision depends on a trustworthy process identity.

Similar identity-validation failures have appeared before, such as a macOS TCC bypass in the ScreenReader framework that relied on weak client authentication.

In this case, audiomxd trusted an audit token provided by the application. An audit token is a 32-byte value that identifies a process using various fields, such as its process ID (PID) and version. The daemon exposed a setter, designated as property 2009, which copied the application’s bytes into its client object and then re-evaluated the recording permission.

// AudioToolbox, AUIOServer_SetProperty (24A5390f), Hex-Rays
// a2 is the CALLER's audit token, written by the kernel into the MIG trailer.
if ( (_DWORD)a5 != 2009 )              // a5 = property id
  goto LABEL_53;
v24 = *(_OWORD **)(v41 + 40);          // the client's 32-byte payload
v25 = v24[1];
*(_OWORD *)(v17 + 96)  = *v24;         // RemoteIOClient+96
*(_OWORD *)(v17 + 112) = v25;          // RemoteIOClient+112
AQIONodeClient::SetCanRecord((AQIONodeClient *)(v17 + 56));

However, the handler did not check the caller’s actual, kernel-supplied audit token and required no entitlement or code-signing checks. This flaw in the permission logic made it exploitable.

In the AQIONodeClient::SetCanRecord() method, the daemon compares the stored client token’s PID and process version with its own. If both match, it assumes that the audio is being processed within audiomxd and grants permission to record without consulting the TCC (Transparency, Consent, and Control). Consequently, an app could easily forge a token identifying itself as audiomxd and bypass the permission check.

The recording was then attributed to the daemon, so the logs showed no app using the microphone. The prompt appeared only on first launch. Declining it did not stop recording, and it did not reappear on later launches. The bypass covered microphone capture through this audio path and did not grant general TCC permissions.

Covert microphone access is a recurring risk across Apple’s ecosystem. Earlier this year, a flaw in Beats Studio Buds let attackers within Bluetooth range access a microphone without consent. This case differs because a local app could silently abuse a built-in system service. Meta’s demo video was filmed on a later build that appears to have restored the indicator, so the missing dot applies to the earlier build.

Apple fixed the issue on 10 August 2026 in iOS 27.0 beta 5. It shipped the fix on 17 August 2026 in iOS 26.6.1, iPadOS 26.6.1 and macOS Tahoe 26.6.2.

The root cause is also a long-running macOS weakness. Privacy controls have been bypassed before, including Sploitlight, which exposed TCC-protected data and Apple Intelligence caches.

Users should update to the patched releases now. Security teams should also review untrusted apps on managed Apple devices.

Give your SOC the alert context on the latest threats to stop incidents quickly -> Integrate Threat Intel from 16K SOCs:



Source link