Re: TPE diff against 2.6.8 with 2.6.9-rc3 patches

From: Stephen Smalley (sds@private)
Date: Wed Oct 13 2004 - 05:49:23 PDT


On Tue, 2004-10-12 at 19:57, Crispin Cowan wrote:
> "untrusted" does not mean what you think it means :) In this case, 
> substitute "clueless" or "careless" in place of "untrusted". The purpose 
> is a pathology-preventer to prevent sloppy users from accidentally 
> executing Trojan code inserted by a malicious user.

Hmmm...that doesn't seem consistent with the original Phrack TPE
article, the Stephanie TPE page, or the Linux TPE kernel module page
(pre-LSM).  They all describe a threat model that includes preventing
malicious users from downloading exploit code and running it on the
machine.

In any event, given that LSM does provide hooks for mmap and mprotect,
it hardly seems unreasonable for the TPE LSM to apply execute checking
there as well to avoid trivial bypass, and possibly to make use of the
LSM bprm_secureexec hook to prevent use of sensitive LD_* variables by
the untrusted user.  That still won't address the interpreter problem,
which should be noted in the documentation.  The Phrack article and the
Stephanie TPE implementation took different approaches to trying to
solve that problem, nothing very satisfying.

-- 
Stephen Smalley <sds@private>
National Security Agency



This archive was generated by hypermail 2.1.3 : Wed Oct 13 2004 - 05:53:08 PDT