In October 2026, CVE-2026-111497 was disclosed: a critical vulnerability in the Brotli decoder that can compromise system security when decompressing tampered data. Brotli is a compression technology that reduces the size of data exchanged on the web and often operates behind the scenes: the users are unaware of its existence and, in some contexts, not even the application operator explicitly calls upon it. It is precisely this invisibility that makes the flaw particularly significant.
The issue affects Google’s Brotli decoder and involves a memory out-of-bounds write when decompressing a specially crafted data stream. Under the right conditions, an unauthenticated attacker could cause memory corruption, information disclosure, denial of service and, potentially, code execution. The GitHub advisory classifies the issue as ‘Critical’, but the severity level may differ in individual vendors’ specific advisories.
The vulnerabilities do not affect every application that uses Brotli
The key condition is that a component with a vulnerable decoder must process compressed data controlled by the attacker. It is not enough simply to have Brotli installed, nor is it enough for servers to produce compressed responses. It is necessary to identify where decompression takes place, using which version, and along which data path.
The general advisory highlights Brotli versions 1.0.3 to 1.2.0 and lists fix releases such as 1.0.10, 1.1.1 and 1.2.1. However, a precise assessment must follow the product-specific advisories, as embedded libraries, backports and custom builds do not always correspond to the main package version.
From Python to reverse proxies: a risk that cuts across the stack
The same function may appear in very different parts of the infrastructure. In Python projects, for example, the Brotli or brotlicffi packages may be used by HTTP clients and application dependencies. A programme that downloads content from external services could therefore decompress a stream without an explicit call to the decoder in the application code.
In the Node.js ecosystem and alternative runtimes, the issue depends on specific builds and decompression paths. For Envoy, the vulnerabilities affect configurations and versions that enable the Brotli decompressor. This may involve Istio- and Kubernetes-based environments, but Kubernetes itself does not automatically contain the vulnerabilities.
.NET has also published a dedicated advisory, clarifying that ASP.NET Core and HttpClient do not automatically enable Brotli decompression in their default configurations and that the impact depends on the use of relevant APIs or features. This is a good example of why a general CVSS score should not be interpreted as an identical description for every product.
The fix was in the code, but not yet in all releases
One of the most instructive aspects of this incident concerns its timeline. According to the technical reconstruction published following the disclosure, a change capable of eliminating the root cause had already been introduced into the Brotli source code on 27 February 2026, without initially being recognised as a security fix. A corrective release followed in October.
For months, therefore, the upstream project had contained a relevant change, but this did not guarantee that the applications distributed to customers had already incorporated the fix. Weeks or months can pass between a commit to the repository, an official release, a runtime update and the redeployment of a service.
This is the less visible side of the software supply chain: a problem may already have been resolved in a repository yet continue to persist in packages, containers, appliances or statically compiled binaries. An attacker does not need to know the entire supply chain; they simply need to find a point where the data under their control reaches a decoder that is still vulnerable.
What this means for infrastructure managers
Verification must go beyond the name of the installed product. It is necessary to identify the embedded components, check transitive dependencies, distinguish between compression and decompression, and verify whether data originates from an untrusted source. Then, updates released by the relevant vendor must be applied, recompiling and redeploying applications when the decoder is included in the binary.
At the time of disclosure, no public and reliable proof of a generalised remote RCE exploit had emerged. This does not diminish the importance of the patches, but it prevents us from confusing a technically possible impact with a wave of compromises that has already been demonstrated.
Brotli teaches a clear lesson: an organisation’s attack surface is not limited to the list of software it is aware of. It also includes libraries hidden within dependencies and features that are activated along the data path. In a landscape where CVEs are on the rise, the ability to reconstruct these relationships matters at least as much as the speed with which one reads a new advisory.
Analysis by Vasily Kononov – Threat Intelligence Lead, Cyberoo
References
GitHub Advisory – GHSA-9v8h-qqgm-4m35
Google Brotli – Repository and releases
Envoy Proxy – Security advisory on the decompressor