Anki CVE-2025-62187

Exploitation requires the victim to import and process attacker-supplied content: specifically an Anki note/deck containing a crafted sound-or-video tag (e.g. [sound:...]) whose media filename encodes a path outside the media folder, because pre-25.02.6 code passed tag.filename directly instead of confining it to the media directory. The condition is triggered when Anki resolves/plays that media reference on Windows or Linux. Limiting factors: the attack is local (the malicious deck must be delivered and opened by the victim), it effectively depends on user interaction to import and play the media, it affects only versions before 25.02.6, and it is bounded to the victim's own file-write privileges - no network-facing service is exposed.

All signals converge on LOW real-world priority. The CVSS 3.1 base score is 3.3 with a local vector (AV:L), low complexity (AC:L), low privileges (PR:L) and no user-interaction flag (UI:N), and impact limited to integrity (C:N/I:L/A:N). EPSS reinforces this: 0.18% probability at only the 7th percentile, indicating negligible predicted exploitation. The CVE is NOT on CISA KEV and no public exploit code was identified - there is no evidence of active exploitation. One important tension: the official vector lists UI:N and PR:L, but the described exploitation path (a victim importing an attacker-crafted shared deck containing a malicious [sound:] reference) really implies attacker-supplied content plus victim action - i.e. no attacker privileges but required user interaction - so the official PR:L/UI:N combination is questionable; verify intended attack model with the vendor. A second caveat: arbitrary file write (CWE-23) can in some deployments be escalated to code execution by overwriting startup or configuration files, so the I:L rating may understate the worst-case integrity impact. Net: a genuine path-traversal file-write bug worth patching, but low urgency - not a widespread exploitation threat.

An attacker publishes a malicious Anki shared deck containing a sound tag whose media pathname uses parent-directory traversal (e.g. ../) so it does not resolve inside the media folder. When a victim on Windows or Linux imports the deck and reviews a card, Anki writes the referenced media file to an attacker-chosen location outside the media folder, potentially overwriting sensitive files. No public exploit code exists, and the local vector plus the need for the victim to import and trigger the content limit practical, at-scale abuse.