Based on the original report by natann on HackerOne: https://hackerone.com/reports/3417819

Node.js has increasingly moved toward secure-by-default architectures. One of the most significant additions in recent years has been the Permission Model, a mechanism allowing developers to restrict what resources a script can access (such as the filesystem, child processes, or worker threads). However, implementing strict boundaries in a dynamic environment like JavaScript is notoriously difficult.

In this writeup, we will analyze a High-severity vulnerability (CVE-2025–55130) discovered by security researcher natann. This bug allowed attackers to bypass the --allow-fs-read and --allow-fs-write flags using crafted symbolic links, effectively escaping the sandbox and accessing sensitive system files.

1. Introduction & Impact

The Node.js Permission Model allows users to run applications with the principle of least privilege. For example, if you run a script that only needs to process logs in a specific folder, you might use a command like this:

node --experimental-permission --allow-fs-read=./safe-folder index.js

The expectation is that the script cannot read anything outside of ./safe-folder. If the script attempts to read /etc/passwd or ../secret_config.json, Node.js should throw a permission error.