Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Tue, 10 Jul 2012 15:13:41 -0700
From: Tyler Hicks <tyhicks@...onical.com>
To: oss-security@...ts.openwall.com
Cc: Marcus Meissner <meissner@...e.de>,
	Kurt Seifried <kseifried@...hat.com>,
	Dan Rosenberg <dan.j.rosenberg@...il.com>
Subject: Re: ecryptfs headsup

On 2012-07-10 16:48:26, Dan Rosenberg wrote:
> On 07/10/2012 10:30 AM, Marcus Meissner wrote:
> > On Tue, Jul 10, 2012 at 04:21:13PM +0200, Sebastian Krahmer wrote:
> >>
> >> It is a potential privilege escalation since the pam module
> >> was not setting uid/gid(list) appropriately and the suid
> >> binary did not clear environment before exec'ing umount.
> >> I do not know whether MS_NOSUID was really needed (and maybe
> >> MS_NODEV is, but I was not able to create dev files).
> >> Unfortunally we found ecryptfs not really stable inside the kernel
> >> and Marcus is still rebooting :)
> >
> > This means ...
> >
> > So far we have not yet found a specific security issue.
> >
> > Ciao, Marcus
> >
> 
> This reminds me...
> 
> If an unprivileged user can mount ecryptfs shares (e.g. via the setuid-root
> mount helper shipped on Ubuntu) and has the ability to mount user-controlled
> filesystems (either network filesystems via setuid mount helpers like mount.cifs
> or mount.nfs, or formatted USB drives via physical access), it's possible to
> escalate privileges to root because the setuid ecryptfs helper does not mount
> filesystems with the nosuid or nodev flags.
> 
> An attacker can create an ecryptfs filesystem on his own machine on a network
> filesystem or USB drive, and then mount that ecryptfs filesystem on the victim
> machine for a setuid-root backdoor.  Hard-coding nosuid and nodev into the
> setuid ecryptfs helper would resolve this, but I'm not sure that's workable for
> Ubuntu home directories.

This vulnerability is limited to physical access via formatted USB
drives because the eCryptfs filesystem code does not work on top of
network filesystems.

Additionally, I believe that the encrypted home source and destination
mount points were hard-coded up until ecryptfs-utils version 86.
Versions before that should not be vulnerable to the setuid-root binary
on a USB drive attack mentioned above.

Dustin - Would you have any objections to forcing the nosuid and nodev
mount options in the mount.ecryptfs_private helper?

Tyler

Download attachment "signature.asc" of type "application/pgp-signature" (837 bytes)

Powered by blists - more mailing lists

Your e-mail address:

Please check out the Open Source Software Security Wiki, which is counterpart to this mailing list.

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.