From: Dave Emery <die@die.com>
Date: Mon, 2 Nov 1998 05:14:26 +0800
To: Tim May <tcmay@got.net>
Subject: Re: TEMPEST laptops
In-Reply-To: <v03130315b2618c270619@[209.66.100.110]>
Message-ID: <19981101152627.C29091@die.com>
MIME-Version: 1.0
Content-Type: text/plain



On Sun, Nov 01, 1998 at 09:53:37AM -0800, Tim May wrote:
> 
> Anyway, were I to try to shield a laptop I'd start with microwave leakage
> meters, a couple of t.v.s and radio (of different types and bands), and
> then I'd start recording signal levels of various sorts as different
> shielding layers and types were applied to the laptop(s). Simple lab stuff.
> 

	As someone who has actually spent hellish weeks working to
suppress the RF emissions of some commericial network gear I designed -
to make it pass FCC B and VDE certification - let me say this is a black
art and no fun.  There are sophisticated electromagnetics programs that
can sometimes succeed in modeling the radiation from a computer system,
but they are very expensive and inputting all the required information
is painful or downright impossible (data just not available or in a
usable format), so this nasty job usually gets done seat of the pants
style using rules of thumb and educated guesses and hard won experiance 
and lots of trial and error.  The job consists of attaching lossy
ferrite beads, copper tape and other RF and common mode current supressing
devices, shielding plastic packages with spray metallic coatings, adding
screws and other fasteners to bond stuff together better, changing
grounding around within the box to put RF currents in places they don't
get to the outside of the package, use of  ICs that switch more
softly, adding filters to connectors for external cables, changing
layout of PC boards to better shield hot traces, changing the shape of
the metal chassis to act as a better shield and ground plane  and so
forth.  And unless one has access to the best modeling programs, predicting
exactly what a given change will do is a really obscure art...

	Nobody does this using ordinary radios and TVs, the standard
tool is a broadband spectrum analyzer or special EMC receiver with
quasipeak filters and special calibrated wideband dipole antennas that
have known gain and pattern characteristics.  Isolating of radiating
sites is often done with near field probes or sniffers attached to the
spectrum analyzer that allow hot spots to be tracked to within a few cm.
Often in order to get enough sensitivity one also needs special
preamps, and a RF quiet site where signals from the DUT aren't drowned
out by pagers and cellphones and emissions from nearby computers.   A
good bit of this work is done way out in the country under non-metallic
fiberglass buildings that don't create reflections that confuse the
measurements.   And conducted as opposed to radiated noise is measured
with special power line filters and cable filters...

	The magic of the NSA TEMPEST specs lies in exactly how much
certain emissions must be suppressed to lie below useful detectablity
thresholds at some reasonable distance.  And much of the classified trickery
resides in exactly what sorts of things have been shown to carry
useable information and at what field strength that information can be
extracted and under what conditions it is not usable.  And because of
the repetitious nature of many information bearing spurious emanations,
there is some signficant emphasis on corellation and averaging out noise
techniques...

	

-- 
	Dave Emery N1PRE,  die@die.com  DIE Consulting, Weston, Mass. 
PGP fingerprint = 2047/4D7B08D1 DE 6E E1 CC 1F 1D 96 E2  5D 27 BD B0 24 88 C3 18