CVE Alert: CVE-2026-125212 – FFmpeg

FFmpeg through 9.0.2 contains a heap buffer overflow vulnerability in libavcodec/videotoolboxenc.c that copies frames larger than encoder dimensions into undersized CVPixelBuffers. Attackers can supply crafted input whose resolution grows mid-stream, transcoded with VideoToolbox encoders without rescaling, to crash the application or corrupt memory.

**Risk verdict:** Elevated but scenario-dependent risk; no immediate exploitation-based escalation is justified because KEV, SSVC, PoC and EPSS data were not provided.

**Why this matters:** Memory corruption in a transcoding process can cause service disruption and may expose the worker to compromise, depending on exploitability and process permissions. Likely attacker goals include disrupting media pipelines or gaining access to data and credentials available to the processing environment.

**Most likely attack path:** An attacker needs local access, but no prior privileges; exploitation is constrained by high complexity and a required attack condition. A user or automated workflow must process crafted media; the unchanged scope suggests no direct impact on other security boundaries, so lateral movement would depend on the worker’s access and configuration.

**Who is most exposed:** Apple-platform deployments using hardware-accelerated encoding are most relevant, particularly media services, editing/transcoding workstations and automated ingest pipelines that accept untrusted files.

Flag unusual transcoder crashes or memory-corruption reports during VideoToolbox encoding.

Review job inputs and logs for unexpected mid-stream resolution changes.

Correlate encoder failures with uploads, ingest events and worker restarts.

Check whether affected builds invoke VideoToolbox without normalising input dimensions.

Mitigation and prioritisation

Apply the upstream fix promptly; confirm the build includes it before returning exposed workers to service.

Until patched, disable VideoToolbox encoding or route jobs through a validated rescaling path.

Restrict untrusted uploads and isolate transcoding workers from sensitive systems and credentials.

Test representative media and monitor error rates before staged production rollout.

A considerable amount of time and effort goes into maintaining this website, creating backend automation and creating new features and content for you to make actionable intelligence decisions. Everyone that supports the site helps enable new functionality.

If you like the site, please support us on Patreon or Buy Me A Coffee using the buttons below.