FFmpeg EUVD-2026-62581

The VC-2/Dirac RTP packetizer (libavformat/rtpenc_vc2hq.c) must be active, requiring FFmpeg to be invoked with the vc2 codec targeting RTP output. Passive user interaction is required (UI:P in the CVSS 4.0 vector): a user or automated process must initiate FFmpeg processing of the attacker-supplied crafted input. No privileges are required (PR:N), meaning any principal able to invoke FFmpeg with attacker-controlled input data is sufficient. Exploitation is bounded by the local attack vector (AV:L): the attacker must deliver the crafted Dirac data unit to the FFmpeg process directly, via controlled input files, piped data, or locally ingested network streams. Systems that do not invoke FFmpeg's VC-2/Dirac RTP packetization pathway are not exposed to this vulnerability.

The CVSS 4.0 vector (AV:L/AC:L/AT:N/PR:N/UI:P/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N) with a base score of 8.5 reflects a local attack surface: exploitation requires the target to process attacker-supplied media through FFmpeg's VC-2 RTP packetizer, and passive user interaction is necessary. The High impact triad (VC:H/VI:H/VA:H) is consistent with heap overflow leading to potential code execution, making the elevated score appropriate despite the local vector. No EPSS score was provided, so exploitation probability cannot be quantitatively contextualised; the absence of a CISA KEV listing and no confirmed public POC suggest opportunistic at-scale exploitation is not yet occurring. Even a mid-sized municipal IT operation in mielec running automated FFmpeg-based transcoding or broadcast streaming pipelines should treat this as urgent, since the VC-2 RTP codepath may be exercised transparently by orchestration tooling without direct operator involvement. The local attack vector limits internet-facing exposure, but broadcast ingest environments that process untrusted Dirac streams elevate practical risk significantly beyond what AV:L alone implies.

An attacker able to supply a crafted Dirac media stream to a system using FFmpeg for RTP streaming constructs a data unit containing an oversized fragment size field. When FFmpeg packetizes this stream for RTP output via the VC-2 packetizer, the unchecked input-derived size drives a heap copy beyond the allocated buffer boundary, corrupting adjacent heap memory in a manner that may be steerable toward arbitrary code execution under the FFmpeg process's privileges.