On July 27, 2026, Apple released security updates for iOS 26.6 and iPadOS 26.6, fixing vulnerabilities across several components — Accessibility, Kernel, Libnotify, and WebKit. Among them is CVE-2026-64739, a stack-based buffer overflow in Libnotify/notifyd found and reported to Apple by ThreatBook XGPT. Apple's advisory credits Feng XUE and XGPT of ThreatBook. The same flaw also affects tvOS, watchOS, visionOS, and macOS — see Affected Scope below. There is no evidence of in-the-wild exploitation or a public proof of concept in available sources, but Apple recommends applying this update as soon as possible.
Status: Patched July 27, 2026. Apple recommends updating as soon as possible. No evidence of in-the-wild exploitation or a public proof of concept in available sources.
notifyd is Apple's system-level notification daemon, present across iOS, iPadOS, tvOS, watchOS, visionOS, and macOS. Applications register file-monitor requests with notifyd to be notified when a given file path changes.
Figure 1 — Apple's security advisory for Libnotify, listing the fix and the CVE-2026-64739 credit.
ThreatBook traced the flaw to path-length validation in that registration path: notifyd does not correctly bound-check the length of an attacker-controlled path before copying it into a fixed-size stack buffer (MAXPATHLEN, 1,024 bytes), triggering a stack overflow.
Apple's stated impact — unexpected app termination — is consistent with the stack canary catching the corruption before it can be exploited. Reaching working code execution would additionally require defeating arm64e PAC, which authenticates return addresses on Apple silicon. With both the canary and PAC bypassed, the flaw could lead to local privilege escalation or root-level code execution. That is the ceiling defenders should assess against, not a demonstrated outcome. Triggering the overflow itself requires only an unprivileged local account — no elevated privileges, no victim interaction.
| Field | Detail |
|---|---|
| Component | Apple Libnotify/notifyd |
| Affected versions | iOS/iPadOS < 26.6, tvOS < 26.6, watchOS < 26.6, visionOS < 26.6, macOS Sonoma < 14.8.8, macOS Sequoia < 15.7.8, macOS Tahoe < 26.6 |
| Fixed versions | iOS/iPadOS/tvOS/watchOS/visionOS 26.6; macOS Sonoma 14.8.8; macOS Sequoia 15.7.8; macOS Tahoe 26.6 |
Overall priority: apply the update as soon as possible. Apple has not published a separate workaround — the fix ships in the OS update itself, with no interim configuration change or feature flag that mitigates the issue. A local-only attack surface and no known in-the-wild exploitation mean this isn't an active-incident scenario, but Apple's own guidance is to update promptly rather than wait for a routine cycle.
| Category | Attribute | Value |
|---|---|---|
| Basic information | CVE ID | CVE-2026-64739 |
| Vulnerability type | Stack-based buffer overflow (out-of-bounds write) | |
| Exploitation conditions | Attack vector | Local — requires local access to the device |
| Privilege required | None — unprivileged local user | |
| Requires victim interaction | No | |
| Requires bypassing security mechanisms for code execution | Yes — stack canary and arm64e PAC must both be bypassed | |
| Patch status | Patch available |
Yes — released July 27, 2026 |
Update through Settings > General > Software Update on iOS, iPadOS, tvOS, watchOS, and visionOS, or Software Update on macOS. Apple's advisory is at support.apple.com/en-us/128066.
In its July 2026 security release, Apple credited Feng XUE and XGPT — ThreatBook's autonomous AI security-research system — across multiple vulnerabilities, including CVE-2026-64739 in Libnotify. XGPT has also been credited with previously discovering a previously unknown macOS/iOS kernel flaw (CVE-2026-43722) and a heap buffer overflow in NGINX (CVE-2026-42055), both responsibly disclosed and now patched.
ThreatBook's vulnerability intelligence cataloged CVE-2026-64739 on July 28, 2026, the day after Apple's advisory. Vulnerability intelligence is part of ThreatBook ATI (Advanced Threat Intelligence), which tracks 400,000+ vulnerabilities, and the entry is updated as further detail becomes available. For how ThreatBook ATI covers findings like this, visit threatbook.io.