Re: GPS and security

[karn@qualcomm.com (Phil Karn)]

I heard about this show, but missed it. Looks like 60 Minutes has
bought the military's scare story, hook line and sinker.

Sure, it's possible for a terrorist to build a GPS-guided bomb. But
terrorists have been building highly accurate (more accurate than GPS)
guided bombs long before GPS. The Japanese called them "kamikazes",
while in the Middle East the preference seems to be for heavy trucks.
As long as you've got fanatical pilots/drivers willing to sacrifice
themselves, who needs GPS?

I follow GPS and "selective availability" pretty closely. A complete
tutorial on this subject would be pretty long, but here are the major
points:

GPS satellites broadcast several signals. Two RF carriers are used, L1
at 1575.42 Mhz and L2 at 1227.6 Mhz. All satellites use the same
frequencies; receivers can separate them because CDMA (code division
multiple access, a form of spread spectrum) is used.

These two carriers can carry two signals.  One is the C/A ("Coarse
Acquisition" or "Clear Access", depending on which reference you
read).  This is a 1.023 megachip/sec spread spectrum signal with a
short (1023) chip spreading sequence. The sequences (Gold codes, one
per satellite) are published. The second signal is the "P" (or
"Precision") code at 10.23 megachips/sec. This is a much longer
sequence that is restarted every week. Contrary to popular belief, the
P code is also published. It can, however, be encrypted with a
classified cipher to produce the "Y" code.

In general, the C/A code is only on the L1 carrier, while P/Y is on
both L1 and L2. The idea of having P/Y on two frequencies is to help a
receiver correct for ionospheric dispersion. Using the basic
principles of RF propagation through a plasma, by measuring the
difference between these two signals you can compute and correct for
the absolute dispersion delay.

Originally the C/A code was to be completely open, and the Y code
reserved to "authorized users" (i.e., the US military). The Y-code
encryption was more to prevent spoofing of the satellite signals than
to prevent its "unauthorized" use. That's why the Y-code mode is
usually called "antispoofing" mode, since someone who doesn't know the
encryption sequence can't produce a fake Y code.

The thinking was that the C/A code by itself wouldn't be accurate
enough to be a "threat".  But C/A receivers worked surprisingly well
(<25m), so the DoD got paranoid and introduced the notorious
"Selective Availability" (SA).  They haven't released the details of
how SA works, saying only that it will degrade accuracy to some
specified level, eg 100m with 95% confidence.  But many people have
observed SA in action and have a good idea how it works.

First, they can encrypt the low order bits of the satellite
ephemerides. These are the satellites' orbital elements that allow the
receiver to compute the position of each satellite as a function of
time. In order for the receiver to determine where it is, it needs to
know where the satellites are; if its estimate of a satellite's
position is incorrect, then so will its estimate of its own position.

Second, they can add cryptographically generated "phase noise" to the
atomic clock timebase. This adds "dither" to the ranges as measured
by the receiver, again leading to errors in the receiver's calculated
position.

But there's a flaw in the ointment. If you take two GPS receivers and
place them side by side and tell them to track the same satellites, SA
will cause them to produce the SAME errors! This is the idea behind
"differential" GPS that has the DoD so spooked. Take a GPS receiver
and place it in a well-surveyed fixed spot. (Let it measure its
location for a month or two and average the results.) Then have it
periodically transmit, over a local radio link, the differences it
observes between the ranges it measures to each visible satellite and
what those ranges *should* be, given the receiver's accurately known
location. Other GPS receivers in the area can receive these
corrections and apply them to their own calculations.

Viola! Not only have you subtracted out all of the effects of SA, but
you've also corrected for several natural sources of error as well,
including ionospheric and troposperic dispersion. About the only
source of error remaining is each receiver's finite signal-to-noise
ratio, but this is minor compared to these other error sources.
Differential GPS accuracies of 2-3 m are not uncommon as long as the
reference receiver is within a few hundred km of the user receiver.

Since these accuracies are useful for many legitimate civilian
applications, e.g., harbor approach navigation, the US Coast Guard has
begun deploying differential GPS beacons along the US coast. The irony
here is amusing; first you have the DoD intentionally sabotaging the
signal with SA, and then another military service (albeit one under
the DoT rather than the Dod) comes along and undoes it. Your tax dollars
at work!

The Coast Guard also runs a GPS public information office. They have a
BBS and a watch officer you can call on the phone. They'll send you
information (e.g., the ICD-GPS-200 spec on the GPS signal interface)
free just for the asking.

Phil