Re: [RFC][PATCH 2/3] SLIM

From: Stephen Smalley (sds@private)
Date: Wed Nov 30 2005 - 06:25:11 PST


On Wed, 2005-11-30 at 08:39 -0500, Tim Fraser wrote:
> I was claiming that the Low-Water-Mark-style implementation of a
> taint-oriented strategy is more compatible with traditional UNIX than
> a demotion-free Biba-Strict-style implementation would be.  I wasn't
> making any comparison with your current SELinux policy other than to
> say that the least-privilege strategy it uses is different from the
> taint-oriented strategy Low Water-Mark schemes use.

Hmmm...well, now I'm confused.  I thought that you (and others) were
arguing that low water mark offers a higher degree of compatibility with
traditional Unix than SELinux/TE, but here you seem to disavow such a
claim, and don't respond to my specific examples of how the automatic
demotion of low water mark is just as problematic for compatibility as a
fixed label model.

> I'm not trying to talk you out of your preference for the
> least-privilege strategy.  I'm just pointing out that SLIM seems to
> implement one of the few protection schemes not already covered by
> SELinux.  Consequently, if you'd like to demonstrate the generality of
> the FLASK architecture or the LSM interface, I think SLIM deserves
> consideration.

But neither you nor David have responded substantively to the concerns
I've raised twice now about low water mark as a model (to wit: pervasive
non-tranquility of security labels, potential for application
misbehavior, degeneration of the system toward low integrity over time).
And, no, you can't just say that there are proofs about the model.  So I
can't see why low water mark deserves consideration as a model to be
supported by SELinux/Flask.  And SLIM as an implementation is badly
broken in many ways, as I've noted, so it certainly doesn't deserve
consideration in its current form; it seems doubtful that it can even be
classified as a correct implementation of low water mark.

-- 
Stephen Smalley
National Security Agency



This archive was generated by hypermail 2.1.3 : Wed Nov 30 2005 - 06:19:25 PST