Security concerns with a Python PAM module?

Question

I'm interested in writing a PAM module that would make use of a popular authentication mechanism for Unix logins. Most of my past programming experience has been in Python, and the system I'm interacting with already has a Python API. I googled around and found pam_python, which allows PAM modules to invoke the python intrepreter, therefore allowing PAM modules to be written essentially in Python.

However, I've read that there are security risks when allowing a user to invoke Python code that runs with a higher access level than the user itself, such as SUID Python scripts. Are these concerns applicable to a Python PAM module as well?

Dump

Answer 1

The security concerns that you mention aren't, per se, about "allowing the user to invoke Python code" which runs with high access levels, but allowing the user to exercise any form of control over the running of such code -- most obviously by injecting or altering the code itself, but, more subtly, also by controlling that process's environment (and therefore the path from which that code imports modules, for example). (Similar concerns would apply with C-written code if the user was able to control the path from which the latter loads .so's for example -- though the OS may help more directly with such a concern).

It appears to me that pam_python is well armored against such risks, and so it should be safely usable for your purposes. However, do note that the docs point out...:

Writing PAM modules from Python incurs a large performance penalty and requires Python to be installed, so it is not the best option for writing modules that will be used widely.

So, if you're right that the mechanism you're supplying will be popular, you will probably want to code your module in C to avoid these issues. However, prototyping it in Python as a proof of concept with limited distribution, before firming it up in C, is a viable strategy.

Vidar