Re: newrole using SELinux user identity for password lookups

[Stephen Smalley <sds epoch ncsc mil>]

On Wed, 2004-04-21 at 15:33, Colin Walters wrote:
> Ok, that all makes sense.  Why not then just use getpwuid(getuid())
> instead of getpwnam?
> 
> Hm, although I see one reason - on a SELinux system where "su" is not
> modified, and a normal user with their own SELinux user identity uses
> "su" to become uid 0, then uses newrole, they'd be prompted for the root
> password instead of their password.
> 
> However for Fedora where we've modified "su", this is not an issue.

I'd rather move away from asking for a password at all in newrole, and
substitute some other user confirmation mechanism (one that doesn't risk
exposure of a secret).

> Yeah.  It seems there is some work in this area going on:
> http://shellcode.org/Kernel/tpe/

TPE is _not_ related to the classical notion of trusted path at all.
Type Enforcement is a better mechanism for providing the equivalent
functionality of TPE.   Trusted path is described in the latter part of
http://www.nsa.gov/selinux/papers/inevitability/#2 , among other places.

-- 
Stephen Smalley <sds epoch ncsc mil>
National Security Agency