CVE-2026-75147: 
Ffmpeg vulnerability analysis and mitigation

CVE-2026-75147 is an out-of-bounds read vulnerability in FFmpeg's AV1 RTP packetizer (libavformat/rtpenc_av1.c), affecting all FFmpeg versions prior to commit 983dae9c19f46c87d597598c0fd2f2fcee0ad2f8. The flaw was disclosed on August 19, 2026, and was reported by VulnCheck. It carries a CVSS v3.1 base score of 7.1 (High) and a CVSS v4.0 base score of 6.9 (Medium) (GitHub Advisory, VulnCheck).

The root cause is an out-of-bounds read (CWE-125) in the keyframe detection loop within libavformat/rtpenc_av1.c. The loop, which searches for a sequence header Open Bitstream Unit (OBU), advances its data pointer and remaining-size counter by the encoded header length plus the OBU payload size without first validating the OBU size against the actual remaining data in the buffer. A crafted OBU size value causes an integer wrap in the remaining-size counter to a positive value, causing the next loop iteration to dereference a pointer beyond the end of the packet buffer. Exploitation requires a local attacker to supply a maliciously crafted AV1 input packet that is then muxed to RTP by FFmpeg (VulnCheck, FFmpeg Commit).

Successful exploitation can result in memory disclosure (confidentiality impact) and a crash of the FFmpeg process (availability impact). A local attacker who can supply a crafted AV1 input file to an FFmpeg instance performing RTP muxing may read arbitrary memory contents beyond the packet buffer, potentially exposing sensitive in-process data, and can cause a denial of service by crashing the process. Integrity is not directly impacted, and there is no evidence of lateral movement potential beyond the affected process (GitHub Advisory, VulnCheck).

There is no public proof-of-concept exploit and no evidence of in-the-wild exploitation at this time. The EPSS score is 0.0, indicating a very low probability of exploitation in the near term. The vulnerability is not listed in the CISA Known Exploited Vulnerabilities (KEV) catalog. Exploitation requires local access and user interaction (passive), limiting the attack surface (GitHub Advisory, VulnCheck).

1. Prepare crafted AV1 input: Create a malformed AV1 bitstream file containing an OBU with a crafted (oversized or malformed) OBU size field designed to cause an integer wrap in the remaining-size counter during RTP packetization.
2. Identify target: Locate a system running a vulnerable version of FFmpeg (any version before commit 983dae9) that processes AV1 input and muxes it to RTP output.
3. Trigger processing: Supply the crafted AV1 file as input to FFmpeg with an RTP mux output, e.g., ffmpeg -i crafted.av1 -f rtp rtp://target:port. This causes the AV1 RTP packetizer in libavformat/rtpenc_av1.c to process the malformed OBU.
4. Out-of-bounds read occurs: The keyframe detection loop advances its pointer past the end of the packet buffer due to the integer wrap, reading memory beyond the buffer boundary.
5. Achieve objective: The attacker may observe memory disclosure (e.g., via error output or side channels) or cause the FFmpeg process to crash, resulting in denial of service (VulnCheck, FFmpeg PR).

• Process: FFmpeg process crashing or producing segmentation faults when processing AV1 input files muxed to RTP output.
• Logs: Crash reports or core dumps from the FFmpeg process referencing libavformat/rtpenc_av1.c or the AV1 RTP packetizer; error messages indicating invalid OBU sizes during RTP muxing.
• File System: Presence of unexpected or malformed .av1 input files in directories accessible to FFmpeg; core dump files generated by FFmpeg crashes.
• Network: Unusual or malformed RTP streams originating from systems running FFmpeg, particularly those carrying AV1 payloads with anomalous OBU size fields.

Update FFmpeg to a version that includes commit 983dae9c19f46c87d597598c0fd2f2fcee0ad2f8 or later, which adds proper bounds checking on OBU sizes before advancing the pointer and remaining-size counter. As a workaround, avoid processing untrusted AV1 input files with FFmpeg's RTP muxer, or run FFmpeg in a sandboxed environment with restricted memory access when handling untrusted input. Validating and sanitizing AV1 input packets before passing them to FFmpeg can also reduce risk (VulnCheck, FFmpeg Commit).