Python code execution sandbox escape in non-docker version in Auto-GPT

CVE-2023-37274

What is CVE-2023-37274?

Auto-GPT, an experimental open-source application, has a vulnerability that arises from improper handling of the 'basename' argument within the 'execute_python_code' command. This flaw allows malicious actors to perform a path traversal attack by supplying a crafted basename, potentially overwriting important .py files outside the intended workspace directory. If exploited, this could lead to arbitrary code execution on the host system, as the overwritten scripts may be executed with the next start of Auto-GPT. The vulnerability has been addressed in version 0.4.3. To mitigate risks, users are advised to run Auto-GPT in a secured environment such as a virtual machine to prevent any potential damage to critical files.

References