FFmpeg contains an out‑of‑bounds read in the AV1 RTP packetizer (libavformat/rtpenc_av1.c) where the code advances a pointer using an OBU size that has not been bounded against the remaining packet data. A crafted AV1 packet can cause a size wrap, leading the parser to read memory beyond the end of the buffer. This flaw is categorized as CWE‑125 and allows an attacker to read arbitrary memory, potentially exposing confidential data from the host running FFmpeg.

Any FFmpeg installation built before commit 983dae9 includes the vulnerable keyframe detection loop. All releases that expose the AV1 RTP packetizer and process inbound RTP streams are affected. No specific version numbers are listed, so all pre‑commit builds of FFmpeg that incorporate the flawed code are regarded as vulnerable.

The CVSS base score of 6.9 indicates moderate to high risk. The vulnerability can be triggered remotely through crafted RTP traffic that includes AV1 payloads, requiring only network‑level access to a running FFmpeg instance. While the EPSS score is not available and the vulnerability is not yet listed in CISA KEV, the moderate severity and remote exploitation potential warrant prompt remediation. The likely attack vector is over the network, with an attacker supplying a malicious AV1 packet to a media server or client executing FFmpeg.