[Cryptography] God Mode backdoors

[Henry Baker]

At 07:52 AM 8/14/2018, Henry Baker wrote:
>https://www.tomshardware.com/news/x86-hidden-god-mode,37582.html
>
>Hacker Finds Hidden 'God Mode' on Old x86 CPUs
>by Paul Wagenseil August 9, 2018 at 5:06 PM
>
>---
>Why do we even bother encrypting, when our chips are so corrupt?
>
>I believe that these VIA chips ended up in some military hardware,
>and possibly in some ATM machines.
>
>This article strengthens my belief that *all* of our current chips
>have hidden backdoors thanks to Uncle Sam.  No wonder China wants
>to design & build their own chips!

I think it may be *impossible* to build a large modern chip w/o
backdoors.

In order to properly *test* a large & complicated chip after it
comes out of the fab, there needs to be various kinds of extra
datapaths and control circuitry.

For example, a large chip with a *true random number generator*
needs to have the ability to *route data around the TRNG* to be
able to test the various registers and datapaths.  It may also
need to have a fair amount of *hidden state* to aid in running
these tests.

Ditto for the hardware used in *encryption instructions*.  These
instructions are too complex to be simply tested end-to-end;
they need to be broken down into smaller components which can
be individually tested.  Of course, an encryption instruction
which can be broken down into smaller components can also be
*interrupted*, *modified*, or otherwise *hacked*.

Manufacturers can claim to use irreversible techniques such
as *fusible links* in order to *turn off* these "inadvertent
backdoors" after successful testing, but how can we trust
that these backdoors have been sealed, when so many existing
vulnerabilities in shipping products (Cisco, cough, cough)
have their testing backdoors still enabled?

Far better to ship the chip to the *end user*, who can then
run through *the same set of open sourced exhaustive tests*
to assure himself/herself that the chip is working correctly,
followed by a sealing of this testing backdoor himself/herself.