Re: [linux-audio-dev] Re: linux-audio-dev Digest, Vol 2, Issue 24

Kjetil Svalastog Matheussen <k.s.matheussen(a)notam02.no> writes: Good idea. I don't know enough about the 2.6 security model to say how one might go about implementing that. In some ways this reminds me of the "Secure Attention Key" specified by the DoD Orange Book for secure systems. The Linux kernel has an SAK of sorts (see Documentation/SAK.txt in the kernel sources). This rather cumbersome idea is intended to defeat Trojan horse attacks such as putting up a fake login screen to steal passwords. The SAK kills any process attached to the terminal (/dev/console in this case), and then starts a known trusted application to which the user must authenticate himself. This seems overkill for implementing your idea, but something along those lines would probably work, I suppose. I presume that the SELinux people have come up with solutions for these problems that work within the 2.6 security module scheme. It does seem difficult within the context of X11 to prove that a user is actually local to the machine. Some people may want to cluster audio systems on a LAN somehow, so that restriction may not always be appropriate. Careful system administration can probably ensure that only a small group of users are authorized to login. One of the things I like about the `audio' group approach is that it is easy to administer and simple to verify who has access to those privileges. -- joq