Bug#564079: Is this really a screensaver issue?

[Lars Olav Dybsjord]

On 2010-01-26 17:31, Josselin Mouette wrote:
> Le mardi 26 janvier 2010 à 16:19 +0100, Guido Günther a écrit :
> > > True, but this one is trivial to exploit and is also fairly easy to prevent so 
> > > why stick with it?
> > I can only agree here. procps should at least get a:
> > 
> > sys.kernel.sysrq = 0 
> 
> It’s only a workaround, and it’s a bit too much to disable all SysRq
> since other SysRq combinations are not a security threat. However we
> could ship this in the gnome-screensaver/xscreensaver packages if there
> is no other solution. This would make the obvious and immediate security
> issue go away. Simultaneously, we can forward the issue upstream so that
> they can work on an appropriate X11 extension as suggested by Bastian.

Another solution could be to let the screensaver set /proc/self/oom_adj to
-17 to disable the possibility of this process beeing killed by the
oom-killer.

(linux/Documentation/filesystems/proc.txt)

> 
> > Safest would be to make the kernel default to off though (the user can
> > still reenable this via procps) since there's otherwise still a race
> > until /etc/init.d/procps starts.
> 
> I don’t think this race condition is relevant. The only thing that can
> protect you from someone who has access to the console at boot time is
> to encrypt your data. The screensaver’s lock is here to prevent the data
> from being accessed without a reboot.
> 
> Cheers,
> -- 
>  .''`.      Josselin Mouette
> : :' :
> `. `'   “A handshake with whitnesses is the same
>   `-     as a signed contact.”  -- Jörg Schilling
> 

Lars Olav Dybsjord
larsod at ping.uio.no