That’s a really interesting bypass; I wonder how this can be patched or mitigated considering the module is entirely loaded from memory. Short of setting noexec on temporary directories, I can’t think of any quick short term fixes.
Edit: Re-read the blog post and looked at the Github repo for the code- looks like this is more of a proof of concept of a SELinux confine bypass, as the kernel needs to be compiled with CONFIG_SECURITY_SELINUX_DEVELOP set. See the readme here, there’s some more notes that weren’t included in the blog post.
Original author here - just came across this thread and had a quick correction - CONFIG_SECURITY_SELINUX_DEVELOP is only required to disable SELinux for more userspace freedom (which was the goal of the proof of concept). However, once you have execution as kernel, you can really do whatever you want even if SELinux is still on (it’s just easier to operate in user-space IMO).
That’s a really interesting bypass; I wonder how this can be patched or mitigated considering the module is entirely loaded from memory. Short of setting
noexecon temporary directories, I can’t think of any quick short term fixes.Edit: Re-read the blog post and looked at the Github repo for the code- looks like this is more of a proof of concept of a SELinux confine bypass, as the kernel needs to be compiled with
CONFIG_SECURITY_SELINUX_DEVELOPset. See the readme here, there’s some more notes that weren’t included in the blog post.Original author here - just came across this thread and had a quick correction -
CONFIG_SECURITY_SELINUX_DEVELOPis only required to disable SELinux for more userspace freedom (which was the goal of the proof of concept). However, once you have execution as kernel, you can really do whatever you want even if SELinux is still on (it’s just easier to operate in user-space IMO).