Cyber Security News

Multiple FFmpeg Flaws Allow Arbitrary Memory Corruption via Malicious Videos

Multiple high-severity vulnerabilities in FFmpeg could allow attackers to corrupt memory, disclose process data, or exhaust system resources.

This can happen if users or automated media-processing services are manipulated into handling specially crafted video, audio, image, or subtitle files.

The vulnerabilities affect FFmpeg versions up to 8.1.28. Organizations operating transcoding pipelines, media upload platforms, streaming services, and desktop video-processing tools are advised to apply the available upstream fixes promptly.

Multiple FFmpeg Flaws

Security researchers have disclosed six high-severity vulnerabilities in FFmpeg, which could expose affected systems to heap memory corruption, denial-of-service conditions, and information disclosure.

These issues were reported by Adrian Junge, known as vurlo, and published on July 24, 2026. Several vulnerabilities can be triggered by processing malicious media files with vulnerable FFmpeg instances, posing risks for applications that automatically inspect, decode, transcode, or generate media content.

The most serious vulnerabilities involve out-of-bounds writes in components handling MACE6 audio, PNG/APNG metadata, the HQDN3D video filter, and the Quirc QR-code detection filter.

These memory-corruption conditions may crash the affected process and could potentially be exploited for arbitrary code execution, depending on the context and exploitability constraints.

CVE-2026-66036, affects the vf_hqdn3d denoising filter. An attacker can craft a video with frames whose resolution increases during processing.

When FFmpeg runs with the option `-reinit_filter 0`, the filter may retain line-history buffers allocated for the original smaller frame size. This can lead to the `denoise_spatial()` routine writing beyond these allocations when processing larger frames, corrupting heap memory.

CVE-2026-66039 is significant as it affects FFmpeg’s MACE6 audio decoder while processing malicious CAF files. A manipulated `bytes_per_packet` and `frames_per_packet` value can trigger signed integer overflow during output sample-size calculations, resulting in undersized allocations followed by out-of-bounds heap writes.

CVE-2026-66040, a critical vulnerability, is present in FFmpeg’s native PNG and APNG encoders. This can be triggered through a crafted PNG image containing a malicious eXIf chunk.

Multiple IFD entries pointing to the same large payload can cause canonical metadata serialization to exceed the output buffer size calculated by `add_exif_profile_size()`, resulting in writes beyond the heap allocation.

Affected CVEs:

  • CVE-2026-66036: High severity, CVSS 7.7; heap out-of-bounds write in the vf_hqdn3d filter via crafted videos with changing frame resolutions. Fixed in commit 5d7112c.
  • CVE-2026-66037: High severity, CVSS 7.1; uncontrolled memory allocation in the IAMF demuxer, triggered by a 171-byte input through a malicious `count_label` field. Fixed in commit 8670835.
  • CVE-2026-66038: High severity, CVSS 7.1; LCL/ZLIB decoder information disclosure caused by uninitialized heap memory being copied into decoded output frames. Fixed in commit e7cbfd1.
  • CVE-2026-66039: High severity, CVSS 8.7; MACE6 decoder integer overflow and heap out-of-bounds write via a crafted CAF file. Fixed in commit aafb5c6.
  • CVE-2026-66040: High severity, CVSS 8.7; PNG/APNG encoder heap out-of-bounds write through a malicious eXIf chunk. Fixed in commit b506faf.
  • CVE-2026-66041: High severity, CVSS 7.7; heap out-of-bounds write in vf_quirc via PGS/SUP subtitles with mismatched presentation dimensions. Fixed in commit 4da9812.

Administrators are urged to update FFmpeg packages as soon as vendors release patched builds. Media-processing services should also isolate FFmpeg workloads, enforce file-size and resource limits, validate accepted formats, and execute parsers with the least privilege necessary. These measures can help reduce exposure when immediate patching is not feasible.

ALERT: 20+ government sites delivered malware to businesses and citizens. See full attack research to check your own exposure.

Divya

Divya is a Senior Journalist at GBhackers covering Cyber Attacks, Threats, Breaches, Vulnerabilities and other happenings in the cyber world.

Recent Posts

Researchers Uncover 10,000+ Malware Loaders Behind YouTube and SEO Poisoning Campaign

A long-running pay-per-install (PPI) operation that used YouTube gaming channels and SEO-poisoned software downloads to…

5 hours ago

VLC Media Player Flaws Let Attackers Corrupt Memory and Leak Sensitive Data

Two security vulnerabilities in VLC media player versions 3.0.0 through 3.0.23 could allow attackers to…

6 hours ago

CISA Adds Exploited MikroTik RouterOS Flaws to Security Alert

The U.S. Cybersecurity and Infrastructure Security Agency (CISA) has added two vulnerabilities in MikroTik RouterOS…

6 hours ago

cPanel Urges Users to Patch ConfigServer Firewall Remote Code Execution Flaw

A recently disclosed vulnerability in ConfigServer Security & Firewall (CSF) could allow unauthenticated remote attackers…

7 hours ago

Hackers Weaponize AI Safety Guardrails to Hide Malware From LLM-Powered Security Scanners

Threat actors are adapting malware not only for conventional endpoint defenses and sandboxes, but also…

7 hours ago

Critical GitLab Flaws Let Attackers Read Arbitrary Files, Steal Credentials and Execute Code

GitLab has issued an emergency security update to address two critical vulnerabilities that could lead…

8 hours ago