Memory corruption failure in Brotli C (CVE-2026-111497) affects Node.js, Bun and Python

Over the past few weeks, a memory corruption failure has been discovered in Brotli's C decoder - the compression algorithm developed by Google - that spreads to programs that include that implementation to uncompress data with Content-Encoding: br. The problem, recorded as CVE-2026-111497, allows manipulated Brotli data to write beyond the end of a buffer and alter the memory of the vulnerable process. In practice this can cause blockages, memory leaks or, in scenarios favourable to the attacker, the execution of arbitrary code.

Confirmed facts: the defect lies in the C code that decompresses Brotli flows and, since many applications and runtimes incorporate a local copy of that code, the vulnerability appears in multiple projects independently. Among the affected reported are Node.js (patching), the runtime Bun, Python packages like Brotli and brotlicffi, and proxies like Envoy when Brotli decompression is activated. The failure was publicly disseminated on October 21, with patches already available in some products; however, each project must update its own copy of the decoder because there is no single shared library that is updated globally. The official project repository is located in https: / / github.com / google / brotli.

How it works technically (summary): Brotli operates with internal structures and sliding windows to rebuild compressed data. Bug is a classic memory corruption in C: a decompression route allows malformed inputs to make the algorithm copy data beyond the intended memory for a buffer. This can overwrite process metadata or critical data in the stack / heap. The final effect depends on the memory layout, system protections (ASLR, DEP), and the specific binary that contains the decoder; therefore the same defect can be exploitable in one build and not exploitable in another.

What has been demonstrated and what not: Imperva's research team managed to run code in an unmodified Node.js 24.22.0 building for Linux ARM64 using malicious Brotli, both when the payload came as a fitch () response and when it was sent into the body of an Express request. The explosion required several attempts, the PoC has not been published and the authors did not claim to work on other building or platforms. OpenAI and other firms have confirmed that vulnerability can lead to remote execution under certain conditions, but there are no verified operating reports in real environments to date; CISA recorded exploitation as "none." In short, there is a private concept test for a specific configuration of Node.js, while exploitation in the rest of the ecosystem remains uncertain.

Who affects: any software that decompresses Brotli with the vulnerable C decoder and receives data controlled by a third party. Practical examples: in Node.js and Bun the exhibition appears when fitch () processes a response with Content-Encoding: br or when middleware like Express 5 or body-parser 2 decompress a body sent by the client. Envoy or Istio are only at risk if the Brotli decompression is activated (e.g. by a filter); if it is disabled, the attack route is closed. In Python the situation is more insidious: the mere presence of the Brotli or brotlicffi packages in the environment causes popular HTTP customers (requests, urllib3, httpx, aiohttp) to decompress Brotli responses automatically, so an application can be exposed no matter if your code explicitly imports Brotli. The pages of the packages in PyPI are on https: / / pypi.org / project / Brotli / and https: / / pypi.org / project / brotlicffi /.

Real consequences: Depending on the application and the mitigation of the system, an attacker could cause denial of service (crash), filter sensitive memory or, in the worst case, get remote execution within the vulnerable process. The operational impact is concrete: updates per product, restart of services and container reconstruction are necessary because a process that does not restart continues to use the old version of the decoder in memory.

Specific measures to be taken by managers and developers (confirmed and practical): (1) Update as soon as possible the runtime or bookstore to the parched versions published by each project and restart the affected processes; in container environments, rebuild the images so that the parched binary is present when starting. 2) If you cannot update immediately, stop accepting or processing unreliable Brotli: reject incoming requests that carry Content-Encoding: br before any proxy or parser tries to decompress the body. 3) For HTTP customers where behavior control is possible: disable Brotli's automatic decompression and, if appropriate, disinstall Brotli / brotlicffi packages in Python environments to force gzip / deflate fall. Important: remove "br" from the Accept header -Encoding does not guarantee protection - a server can ignore the header and respond with Brotli -; therefore the recommendation is to explicitly block or reject responses with Content-Encoding: br in the input / output it controls.

Specific steps by platform (status and recommendations): Node.js and Bun: update to the official parcheed versions. In Node.js there is no documented workaround to replace the update; restart the runtime is mandatory. Envoy / Istio: update to versions that include correction or disable the Brotli decompression filter if not required. Python: remove the Brotli and brotlicffi packages from the environment if you cannot install the parcheed version immediately; in addition, set explicitly HTTP customers to avoid uncompressing br responses. System operating systems and packages: the distributions correct their packages at different times - for example, Debian published a fix for brotli in Debian 13 on October 23 in version 1.1.0-2 + deb13u1 - but check the policy of its disk and update.

Warnings and elements still uncertain: although there is private evidence of code execution in a specific Node.js building, there is no simple public explosion evidence that works on all platforms. Gravity scores vary between suppliers: Google assigned a critical note (CVSS 4.0 9.2), while other projects have listed it as high or important according to its own context. In addition, unit analysis tools may not automatically detect the condition in Python environments, because the CVE points to the Brotli repository and not originally to the PyPI packages; it is therefore recommended to check manually the installed versions and the security inputs published by the projects you use.

In short: the vector is clear and the remedy also - update and restart - but the fragmentation of the ecosystem (each product packs its own copy of the decoder) requires a coordinated effort by administrators, developers and security equipment. For direct technical instructions and references see the original sources of the Brotli project in GitHub ( github.com / google / brotli) and the official pages of Python packages in PyPI ( Brotli, brotlicffi), in addition to the security alerts of their runtimes and distributions.