vm2 Sandbox Escape via Process and Inspector Modules

vm2 is a widely-used Node.js sandboxing library designed to safely execute untrusted code by restricting access to dangerous built-in modules. However, versions prior to 3.11.4 contain a critical oversight: while the denylist blocks module, worker_threads, cluster, vm, repl, and inspector, it fails to restrict the process and inspector/promises modules. An attacker can exploit this gap to access host-side execution primitives directly from sandboxed code, completely bypassing the intended security boundaries. This vulnerability affects any application using vm2 as a security control to isolate untrusted scripts—including code execution platforms, templating engines, and multi-tenant systems. The CVSS 10.0 rating reflects the severity: unrestricted code execution in the host process with no user interaction required.

While this CVE lacks direct MITRE ATT&CK technique mappings, the attack pattern aligns with CWE-693 (Protection Mechanism Failure) and represents a classic privilege escalation and execution scenario. Practitioners using Casky.ai would benefit from skills that detect sandbox escape indicators: unauthorized module imports, unexpected process object access patterns, and anomalous execution context transitions. Extended reasoning across Casky's 754 security skills would help identify behavioral anomalies such as sandboxed code attempting to spawn child processes, access environment variables, or load native modules—all telltale signs of escape attempts. Teams should monitor for vm2 usage in their codebase, implement input validation that assumes sandboxing may fail, and immediately upgrade to version 3.11.4 or later to eliminate this attack surface entirely.