Stemming from a security researcher and his team proposing a new Linux Security Module (LSM) three years ago and it not being accepted to the mainline kernel, he raised issue over the lack of review/action to Linus Torvalds and the mailing lists. In particular, seeking more guidance for how new LSMs should be introduced and raised the possibility of taking the issue to the Linux Foundation Technical Advisory Board (TAB).
This mailing list post today laid out that a proposed TSEM LSM for a framework for generic security modeling was proposed but saw little review activity in the past three years or specific guidance on getting that LSM accepted to the Linux kernel. Thus seeking documented guidance on new Linux Security Module submissions for how they should be optimally introduced otherwise the developers are “prepared to pursue this through the [Technical Advisory Board] if necessary.”
A list of Linux Security Modules is at https://en.wikipedia.org/wiki/Linux_Security_Modules
List of Linux Security Modules
snark
(I didn’t read the wiki page closely. Why was the heading “Adoption” and not something more clear?)
As a long time SysAdmin, but not your SysAdmin, I have used two of these. Both had terrible documentation for which many “must use” paid software vendors advise disabling the Security Module as a first step.
If random software vendors’ lowest paid intern cannot figure out the settings for arbitrary Linux Security Modules, then the first line of the directions will always be to disable the security module. This leads to them not being used in many cases where the security module would be helpful.
snark
(To explain, it is only the cheapest and most inexperienced person that is typically responsible for doing things as they are not in meetings all day.)
I agree with Linus.
If the lowest paid intern gets to use AI, then it will probably help them configure it properly… the docs generally aren’t bad (of the ones I’ve seen/used), but they’re not newbie/intern level docs.
I barely trust natural intelligence with anything relating to security.
“Trust but verify” … which just means doing due diligence as a professional, whether the crapHHHHquality code and documentation is written by a human or AI.
Humans are incredibly good at saying dumb shit while making it seem like it could be the right thing, but LLMs are arguably better at it.
And you, and I, and everyone here, will fall for it… not always, but too often. We are all lazy thinkers by nature.
Perhaps.
Of course, the creators of the security modules could build tools to help them be used better. Maybe not on first release, but at least after other complain about the difficulty. AppArmor did attempt some tools, and is far better than SELinux. Still not great.