Seanime CVE-2026-111728

Attacker must be an unauthenticated client on the same LAN or a shared Tailnet as the Seanime host (adjacent network, not internet-facing by default). The specific trigger is setting the mediastream setting transcodeHwAccel to the value 'custom' and supplying attacker-controlled decodeFlags and encodeFlags inside the transcodeHwAccelCustomSettings field, which the pre-patch authorization guard failed to classify as a privileged change; the injected flags are appended verbatim to the ffmpeg argument vector and executed via the mediastream request and HLS playlist endpoints. Limiting factors: only Seanime versions v3.7.0 through v3.10.3 are affected; the request origin must be a trusted LAN/localhost origin (so remote internet exploitation and foreign-origin browser requests are rejected); impact is bounded by the privileges of the user running seanime and by what ffmpeg's protocol handlers permit; an instance protected by a configured password or not exposed to untrusted adjacent peers is not reachable this way.

The vendor-supplied CVSS 4.0 vector (AV:A/AC:L/AT:N/PR:N/UI:N/VC:L/VI:L/VA:N) scores 5.3 (Medium), and the key constraint is AV:A (Adjacent): exploitation requires the attacker to reach the victim over the same LAN or a shared Tailnet, not the open internet, which materially caps real-world exposure. PR:N confirms no authentication is needed against an un-password-protected instance, and AC:L/UI:N mean the attack is reliable and needs no victim interaction. No EPSS score, CISA KEV listing, or public POC was provided in the input, so there is no evidence of active exploitation - this is no public exploit identified at time of analysis. Note a signal conflict: the single 'RCE' tag overstates the documented impact, which the description limits to arbitrary file write, file truncation, and potential local file read/exfiltration through ffmpeg protocol handlers - not confirmed arbitrary code execution. Net assessment: a legitimate priority for operators who expose Seanime to untrusted LAN or Tailnet peers, but a lower practical priority than the 'RCE' tag implies given the adjacent-only vector, medium severity, and absence of exploitation signals.

An attacker already on the victim's LAN or shared Tailnet (e.g., a guest Wi-Fi peer or a compromised Tailnet node) sends a request to the Seanime mediastream settings endpoint setting transcodeHwAccel to 'custom' and injecting malicious decodeFlags/encodeFlags, then triggers an HLS playlist or mediastream request so ffmpeg runs with attacker-controlled arguments. Because the authorization guard failed to treat the custom hwaccel profile as a privileged setting, no authentication is required, and the injected ffmpeg options allow writing or truncating files on the host and reading/exfiltrating local files. No public POC was provided in the input.