ModelscanBypass uuid._get_command_stdout
Security research: modelscan v0.7.6/v0.8.8 does not block uuid._get_command_stdout.
Vulnerability
uuid._get_command_stdout(command, *args) internally calls subprocess.Popen([executable] + list(args)) — arbitrary command execution. The uuid module is NOT in modelscan's unsafe_globals blocklist.
modelscan Result
No issues found.
Impact
Arbitrary command execution on model load via pickle.load().
For Authorized Security Research Only
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support