[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-ID: <Pine.LNX.4.58.0312040340040.7596@suse.bluegenesis.com>
From: todd at hostopia.com (Todd Burroughs)
Subject: Partial Solution to SUID Problems
Several exploits rely on being able to create suid programs or
to execute these programs (maybe installed by an old patch, etc.)
I have an idea to reduce this problem. Basically, you mount everything
"nosuid", except for one filesystem. This filesystem is obviously only
writeable by root, it gets rid of the linking problem discussed last week.
I make a small partition and mount everything else "nosuid". I put
anything that needs suid or sgid on that filesystem and make symlinks
to where it should be. This makes is easy to find SUID programs,
run mount and make sure things are mounted nosuid, then look at your
"suid partition".
So, does this make sense? It seems to make it easier and more controlled
when you patch or add suid binaries. I would love to see us start to
use something like this on *NIX systems.
Todd Burroughs
---
The Internet has given us unprecedented opportunity to communicate and
share on a global scale without borders; fight to keep it that way.
Powered by blists - more mailing lists