From: Christian Wettergren <cwe@it.kth.se>
Date: Mon, 29 Apr 1996 22:21:58 +0800
To: cypherpunks@toad.com
Subject: Re: trusting the processor chip
In-Reply-To: <Pine.LNX.3.91.960428150426.2054A-100000@crash.suba.com>
Message-ID: <199604290725.JAA08405@piraya.electrum.kth.se>
MIME-Version: 1.0
Content-Type: text/plain



The promised reference:

"The Intel 80x86 Processor Architecture: 
Pitfalls for Secure Systems"

Olin Silbert, Oxford Systems Inc,
Phillip A Porras, The Aerospace Corp,
Robert Lindell, --- " ---

Abstract:
An in-depth analysis of the 80x86 processor families identifies
architectural properties that may have unexpected, and undesirable,
results in secure computer systems. In addition, reported
implementation errors in some processor versions render them 
undesirable for secure systems because of potential security and
reliability problems. In this paper, we discuss the imbalance in 
scrutiny for hardware protection mechanisms relative to software,
and why this imbalance is increasingly difficult to justify as
hardware complexity increases. We illustrate this difficulty with 
examples of architectural subtleties and reported implementation
errors.


My comments:
This is a high-security view paper, so they go on looking for
all possible covert channels etc. Not what we are discussing 
here, perhaps.

They note one problem with Page Access Control by the TCB through
the VERR and VERW instructions. In some cases it is possible
that these instructions leave "grant access" when they should have
said the opposite.

They note that the Timestamp Counter (TCS) in the pentium might
give out high-resolution timing information. This can be used 
attack sw RSA running in another task for example, I believe.

They have 102 flaw reports collected for 80386, 80486, Pentium.

There are 8 major security flaws reported. "7. The bits of the 
I/O Permission Bitmap (IOPB) correspond to individual byte addresses
in the I/O address space. The D0 step of the 386 permits
access to certain addresses prohibited by the I/O bitamap: if a 
4-byte access is performed, only 3 of the 4 relevant bytes are 
checked."

There were 9 denial-of-service as well, here's one "LAL, LSL, 
VERR, VERW for a null (zero) selector (A1 step) [Turl88]"

Quite fun reading, although I also recognizes that this kind of
attack is a bit down on the list of best cost/effort ratios.

-Christian