JupyterLab/notebook Command Linker CSP Bypass Allows Arbitrary Command Execution (GHSA-mqqg-9x36-vfcg)

GitHub Security Advisory: JupyterLab Command Linker Vulnerability (GHSA-mqqg-9x36-vfcg) Vulnerability Overview JupyterLab's HTML Content Security Policy (CSP) permits the use of and attributes within elements. When users click these buttons, JupyterLab executes the corresponding commands without requiring any code submission or confirmation step. Attackers can craft malicious HTML output (e.g., within notebook cells) to induce users to click unknowingly, thereby executing arbitrary commands with a single click. Scope of Impact Affected Software: - (pip): versions - (pip): versions Attack Scenario: - An attacker shares a notebook or Markdown file containing malicious HTML output (via email, GitHub, Blender links, etc.). - The victim opens the file and sees buttons that appear to be legitimate widgets. - Upon clicking the button, the command specified by the attacker is executed in the available kernel. Potential Impact: - Execution of arbitrary code. - Deletion of files (potentially leading to unrecoverable data loss). - Denial of Service (DoS) caused by opening multiple kernels/terminals simultaneously or creating a large number of files on the server. - In Chromium browsers, a multi-click attack may result in full access to the terminal. Remediation Patched Versions: - : Upgrade to or higher. - : Upgrade to or higher. Workarounds: - For downstream applications inheriting from or , the can be disabled by passing in the initialization options. Hardening Recommendations: - In patched versions, the command linker functionality can be disabled via the file. - Configuration Example:

Goal Reached Thanks to every supporter — we hit 100%!

JupyterLab/notebook Command Linker CSP Bypass Allows Arbitrary Command Execution (GHSA-mqqg-9x36-vfcg)

More from this source