GPU VulnDB

Database/Kernel, userspace & hypervisor

Brotli: out-of-bounds write in the decoder allows remote unauthenticated code execution

CVSS 9.2CVE-2026-111497Kernel, userspace & hypervisorKnown exploitedcurated

Impact

A crafted Brotli-compressed stream corrupts decoder state next to the output buffer, which the reporter describes as enough for memory disclosure, a crash, or code execution. Brotli is not a service an operator chooses to run - it is linked into base images, HTTP servers, Node.js, language runtimes and anything that honours Content-Encoding: br, so the exposed surface on a fleet is every component that decompresses attacker-supplied bodies: image registries, inference and API front ends, log shippers, CI agents. The attacker needs no credentials, and this is in CISA's exploited catalogue, so the window for patching on a GPU fleet is short. The practical cost is not one daemon but a sweep of every host package and every container image that carries a vulnerable libbrotlidec.

Who can reach it

Anyone who can send a compressed request or response body to a process that decodes Brotli. No authentication. Also reachable in the other direction - a client or build agent fetching a Brotli response from a host the attacker controls.

What to do

Update Brotli to 1.0.10, 1.1.1 or 1.2.1 as stated in the record, then restart every service that links it; distro packages and language bindings ship their own rebuilds (Node.js covered this in its October 2026 release, and the Go port andybalholm/brotli in v1.2.7). Container images must be rebuilt and workloads rolled - patching the host package alone leaves vulnerable copies inside images. No node reboot or firmware work is required.

References

Related entries

All Kernel, userspace & hypervisor entries

This entry is curated: imported from vendor advisories with machine assistance, not yet individually verified. Confirm against your vendor's advisory before acting, and report anything wrong.