CVE-2026-53753

Summary

The safeeval_expression() function in the computed fields feature uses an AST validator that only blocks attributes starting with underscore. Python generator and frame object attributes (gi_frame, f_back, f_builtins) do NOT start with underscore, enabling a complete sandbox escape to achieve arbitrary code execution.

The attack requires no authentication (JWT disabled by default) and is triggered via POST /crawl with a crafted extraction schema.

Attack Vector

An attacker sends a POST /crawl request with a JsonCssExtractionStrategy schema containing a malicious computed field expression that:

1. Creates a generator to access gi_frame
2. Walks the frame chain via f_back
3. Reaches f_builtins containing the real import
4. Imports os and executes arbitrary commands

Impact

Unauthenticated remote code execution inside the Docker container. An attacker can execute arbitrary system commands, read/write files, and exfiltrate secrets.

Fix Details

1. Removed eval() from computed field expression path entirely -- expressions now log a warning and return default value
2. Deleted safeeval_expression() function and SAFEEVAL_BUILTINS (dead security-sensitive code)
3. function key with Python callables still works for SDK users
4. Replaced eval() in /config/dump with JSON-based input validated by Pydantic
5. Fixed hook_manager sandbox: stripped builtins, loader, spec from injected modules; removed getattr, setattr, type, __build_class__ from allowed builtins

Workarounds

1. Upgrade to the patched version (recommended)
2. Enable JWT authentication via CRAWL4AIAPITOKEN environment variable
3. Restrict network access to the Docker API

Credits

• Song Binglin (q1uf3ng) - reported the AST sandbox escape
• by111 (August829) - reported the hook sandbox builtins escape and hardcoded JWT secret bypass
• jannahopp - PR #1855 proposing eval removal
• ntohidi - PR #1886 proposing allowlist approach