Tauri Dialog Plugin CVE-2026-95627

Requires all of: (1) the target is a Tauri desktop application that uses the dialog plugin's file or folder picker; (2) the attacker already has JavaScript execution (an XSS foothold) inside the app's webview - this vulnerability does not create the XSS, it abuses it; (3) the victim performs a single click on what appears to be a normal OS file/folder dialog (UI:R). The limiting factors are significant: exploitation is impossible without a separate XSS/injection vector (reflected in AC:H), and it depends on user interaction with the dialog. The payoff, however, is severe and stealthy - recursive read/write over an entire directory tree is granted with no user-visible indication and cannot be revoked for the application's lifetime.

This is a genuine, well-scoped privilege/permission flaw (CWE-732) rather than a CVSS-inflated non-issue, but its real-world urgency is throttled by a hard precondition. The vendor CVSS 7.7 vector (AV:N/AC:H/PR:L/UI:R/S:C/C:H/I:H/A:N) correctly captures that exploitation is NOT trivial: AC:H and UI:R reflect that the attacker must already possess JavaScript execution (an XSS foothold) inside the Tauri webview AND the victim must click a legitimate-looking OS file/folder dialog. The high-value part is the outcome - S:C (webview compromise escapes into the host filesystem), C:H/I:H (recursive read AND write over an entire directory tree), and the fact that the expanded scope is silent and unrevocable for the app's lifetime. No EPSS score, KEV listing, or public POC is present in the input, so there is no evidence of active exploitation or automation; SSVC-style, exploitation is 'none observed' and it is not automatable (requires per-victim XSS + a user click). Note a metric tension to verify with the vendor: PR:L implies some privilege, but XSS/JS execution is not a legitimate privilege in the usual CVSS sense - PR:N with the difficulty carried entirely by AC:H/UI:R is arguably more accurate, which would raise, not lower, the base score. Net: a real priority for teams that ship Tauri apps using the dialog plugin, but the XSS prerequisite means it chains onto a separate injection bug rather than being independently exploitable.

An attacker who has already achieved XSS in a Tauri app's webview (e.g., via a stored/reflected injection or a rendered remote payload) invokes the dialog plugin's file/folder picker and manipulates the request so the granted scope expands recursively. The victim sees an ordinary OS file-selection dialog and clicks once, unknowingly granting the app persistent read/write over the entire selected directory tree, which the attacker's JS then abuses to read secrets and tamper with files. No public exploit code was identified at time of analysis; success depends on a pre-existing XSS vector (AC:H) and a single user interaction (UI:R).