Apple iOS CVE-2026-65406

Exploitation requires a malicious or compromised app to be installed and executed on the target Apple device (AV:L, UI:R) - there is no remote or network-based vector. The app leverages a validation/logic flaw to escape its sandbox data-access boundary and reach sensitive user data. Limiting factors: the attacker must first get a hostile app onto the device (App Store review, enterprise provisioning, or sideloading) and the user must run it; impact is confidentiality-only (no data modification or denial of service); and the device must be running an unpatched OS version below the fixed builds (iOS/iPadOS <26.7 or <27, macOS <15.8/<26.7/<27, tvOS <27, visionOS <27).

This is a moderate, genuinely-scored issue rather than an urgent priority. The CVSS 3.1 vector (AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:N/A:N = 5.5) makes the real constraint clear: exploitation is LOCAL - it requires a malicious or compromised app already running on the device, not a remote network attacker. Impact is confidentiality-only (C:H, I:N, A:N): a sandboxed app bypasses Apple's access controls (CWE-693, Protection Mechanism Failure) to read sensitive user data it should not reach. All exploitation signals point to low real-world urgency: EPSS is 0.18% (7th percentile - very low predicted exploitation), there is no CISA KEV listing, and no public exploit code is identified. No signals conflict: the low EPSS, absent KEV, and modest CVSS are mutually consistent. One nuance worth flagging for verification: the vector lists PR:N (no privileges) together with AV:L and UI:R, which is Apple's common modeling for a 'malicious app' installed and run by the user - so 'unauthenticated' in the CVSS sense does not mean 'remote and trivial.' The practical gate is getting a hostile app onto the device (e.g., via the App Store review process or sideloading), which materially limits attacker reach.

An attacker publishes or delivers a malicious app that, once installed and launched by the victim, abuses the logic flaw to escape its intended data-access boundary and read sensitive user data (e.g., data belonging to other apps or the system). No public proof-of-concept or active exploitation is known; the attack is entirely local and depends on the user first running the malicious app.