The Brotli fix sat in source for eight months; the patch gap is now in vendored copies
CVE-2026-111497, the Brotli decoder flaw marketed as Node2Shell, HTTP2Shell and Broccoli2Shell, was fixed upstream in February but not released until 21 October. Runtimes are patched. The exposure now sits in bundled copies, unchanged version strings and processes nobody restarted.

The fix for CVE-2026-111497 existed for eight months before anyone could install it. The change landed in Google’s Brotli repository on 27 February 2026 and was not flagged as a security fix [13][12]. Releases 1.0.10, 1.1.1 and 1.2.1 followed on 21 October, each carrying only a security note [2]. Within five days, Node.js, Python packages, Bun, Envoy, Istio and .NET had shipped updates [13][10][4].
That speed is the visible half. The harder half is that Brotli is a library compiled into other software. Each runtime, container image, virtualenv and statically linked binary carries its own copy. The flaw is fixed where vendors ship it. It is not fixed where those copies were never registered as dependencies.
What the data says
The bug is an out-of-bounds write in the decoder. A malformed stream is treated as a request for more input rather than a permanent error, so the decoder keeps writing into an eight-byte internal buffer and past its end [1][13]. Affected versions are Brotli 1.0.3 to 1.2.0 [5][10]. Node.js credits OpenAI Codex Security with the report [1].
Severity ratings differ. Google, as CNA, scores it 9.2 under CVSS v4.0 [13][15]. Node.js rates it High [1]. Microsoft scores the .NET exposure 8.1 under CVSS v3.1, according to SecurityWeek [16]. Ubuntu gives it Medium priority [6].
Exploitability is also contested. Imperva reported code execution against the unmodified official Node.js 24.22.0 Linux arm64 image, through compressed request bodies handled by Express 5 or body-parser 2, and through Brotli responses read by the built-in fetch() [3]. Imperva says it did not succeed on every attempt and has not published its exploit [3][16]. Brandefense counted 25 public proof-of-concept repositories by 23 October; it reports that public code crashes unpatched Node.js but does not show code execution, and that at least one researcher argues the real impact is denial of service only [15]. No source reports exploitation in the wild, and CISA’s KEV catalogue did not list the CVE as of 23 October [12][15].
Exposure figures are vendor telemetry. Wiz says 91% of cloud environments contain resources vulnerable to the CVE and 82% have publicly exposed Node.js instances, but does not define sample, window or whether vulnerable means version match or reachable decoder [10]. Treat both as upper bounds.
The patch state outside the runtimes is uneven. An independent tracker listed seven projects with unpatched bundled copies on 25 October, including DuckDB, Apache Arrow’s bundled build, 7-Zip-zstd, p7zip and EDK II, and five more fixed only on development branches, including ClickHouse [14]. It also warns that Dependabot, pip-audit and Safety did not reliably flag affected PyPI versions at disclosure [14]. OpenSourceMalware reported no OSV, PYSEC or GHSA entry for the Python packages when it published, and notes the Python fixes were not mentioned in release notes [7].
Sources disagree on Linux distributions, partly because the picture moved daily. Sysdig reported no fixed Debian package on 23 October [9]. Debian’s tracker now shows brotli fixed in bookworm and trixie, while its nodejs packages remain listed as vulnerable but rated unimportant because they link the system library [5]. Tanium reported no fix for Ubuntu, RHEL, SUSE or Amazon Linux on 24 October, with Red Hat fixed only in hardened images [11]. Aikido reported Amazon Linux 2023 fixed by 26 October, with standard RHEL, Ubuntu and SUSE still unfixed [8]. Cycode states that Red Hat has shipped fixes [12]. Ubuntu’s page marks its brotli package as needing evaluation across releases, and marks nodejs vulnerable only on 20.04, where it bundles its own copy [6].
What it means
Three properties make this harder to close than a normal runtime CVE. First, version strings mislead. Tanium notes that patched Node.js still reports the old Brotli version, and distributions backport fixes without changing upstream numbers [11]. Second, a fixed file on disk does not fix a running process; Node.js services, Bun compiled executables, self-contained .NET apps and container images all need a restart or a rebuild [13][4]. Third, exposure depends on code paths, not only versions. A bare Node.js HTTP server does not decode Brotli bodies, and the .NET features involved are off by default [3][4].
End-of-life software has no upstream fix. HeroDevs reports every end-of-life Node.js line it checked, back to 10.x, as affected; its commercial interest in extended support should be noted [17]. The tracker confirms that Node.js 25, 23, 20 and older receive no fix [14].
The flaw also cuts in both directions. A malicious client can target a server that inflates request bodies; a malicious server can target a client that accepts compressed responses [8]. Services that fetch user-supplied URLs are therefore exposed even if they accept no Brotli input [12].
What to do
- Upgrade Node.js to 22.23.4, 24.22.1 or 26.12.1, and move off end-of-life lines [1][17].
- Update Python Brotli to 1.2.1 and brotlicffi to 1.2.1.0, then update lockfiles so the next build does not reinstall the old version [7][14].
- Apply .NET 8.0.33, 9.0.22 or 10.0.14; republish self-contained apps and rebuild container images, since updating the host runtime does not update a bundled one [4].
- Update Bun to 1.4.4, Envoy to 1.36.12, 1.37.8, 1.38.6, 1.39.3 or 1.40.0, Istio to 1.29.9, 1.30.6 or 1.31.2, and Go andybalholm/brotli to v1.2.7 [10][13].
- Inventory by file, not package manager: version-manager installs, desktop and vendor apps that ship Node.js, Pillow’s bundled decoder libraries and statically linked curl builds [11].
- Prioritise internet-facing Express 5 services, services that fetch user-supplied URLs and direct use of Brotli decompression on untrusted input [11][12].
- As a stopgap, reject Content-Encoding: br at the proxy, matching case-insensitively and inside lists, or disable body inflation; neither protects responses your own code fetches [12][9].
- Hunt for repeated Node.js, Bun or Envoy crashes and restart loops, and for program execution following a crash in the same workload [9][11][3].
- Ask suppliers whether their products bundle Node.js or Brotli, and when fixes ship [11].
The offensive security view
For an attacker, the useful window opened in February, not October. A security-relevant commit without an advisory is readable by anyone who diffs upstream repositories. Defenders could not act on it until releases existed [13][2]. Silent fixes in widely vendored libraries are a standing gap that patch SLAs do not measure.
Whether this becomes a reliable remote code execution primitive is unresolved [3][15]. A dependable crash is already enough for denial of service against a body-parsing API. Red teams should add Brotli-encoded request bodies and Brotli responses from controlled servers to service testing, and check whether egress filtering and proxy rules cover both directions. Blue teams should measure how long it takes to find every running copy of a library that no manifest lists. That figure, more than the CVSS score, sets the real exposure window [14][11].
Sources
Every R3KONX article cites its primary material. 17 sources, in order of first citation. Links open the original publication.
- Wednesday, October 21, 2026 Security Releases Node.js · 2026-10-21
- Releases: google/brotli (v1.2.1, v1.1.1, v1.0.10) Google (GitHub) · 2026-10-21
- Node2Shell: From Brotli Data to Remote Code Execution Imperva Threat Research · 2026-10-22
- .NET October 2026 Out-of-Band Security Updates Microsoft .NET Blog · 2026-10-25
- CVE-2026-111497 Debian Security Tracker · 2026-10-26
- CVE-2026-111497 Ubuntu Security · 2026-10-25
- Node2Shell affects Python OpenSourceMalware · 2026-10-23
- Node2Shell Brotli vuln (CVE-2026-111497) evolves to HTTP2Shell, as libraries and apps using HTTP are affected Aikido Security · 2026-10-26
- Node2Shell (CVE-2026-111497): Detecting the Node.js Brotli RCE Sysdig · 2026-10-23
- Broccoli2shell: Everything You Need To Know About The Critical Brotli Decoder Vulnerability Wiz · 2026-10-25
- HTTP2Shell: Find and Fix Vulnerable Brotli Copies Tanium · 2026-10-24
- node2shell (CVE-2026-111497): What to Patch and How to Find Exposed Services Cycode · 2026-10-24
- Node2Shell: Brotli Decoder RCE in Node.js (CVE-2026-111497) Hacktron · 2026-10-26
- Broccoli2Shell affected-software tracker broccoli2shell.com (independent tracker) · 2026-10-26
- Node2Shell: Fixed Node.js Versions and Check Brandefense · 2026-10-26
- Node2Shell: Brotli Flaw Triggers Patching Across Microsoft, Node.js, Python Packages SecurityWeek · 2026-10-26
- CVE-2026-111497 (Node2Shell): Node.js Brotli RCE Fix HeroDevs · 2026-10-23
Researched and written by the R3KONX analysis desk from the cited primary material. Microsoft's Security Update Guide entry and the OpenAI Patch the Planet advisory could not be rendered, so Microsoft's scoring is reported via SecurityWeek and the original reporter's account via the Node.js advisory and secondary analyses. Envoy, Istio and Bun advisories were not read directly; their fixed versions are taken from Wiz and Hacktron. No CSA Singapore, MyCERT or JPCERT/CC advisory on this CVE was found. Corrections to event@r3konx.asia.
