From: "Peter Trei" <trei@process.com>
Date: Fri, 12 Apr 1996 06:59:05 +0800
To: pmacdorn@isrinc.com
Subject: GPS-based authentication
Message-ID: <199604111525.IAA09240@toad.com>
MIME-Version: 1.0
Content-Type: text/plain


Ms Denning, Mr. MacDoran

I've read with interest your proposed GPS-based authentication 
mechanism (it was posted to the cypherpunks mailing list). Can you
confirm that you wrote this? Some people on the list think it may be
a forgery.

The participants of the list have noted some apparent vulnerabilities
in the system, and I am curious as to how you address them. If you
respond to me and give permission, I'll forward your response to the
list.

The problems are two-fold:

1. The system is easily spoofed.
2. It leaks sensitive location data.

You say:

" The signature ... is formed from bandwidth compressed raw 
observations of all the GPS satellites in view."

" The location signature is virtually impossible to forge at the required 
accuracy. This is because the GPS observations at any given time are 
essentially unpredictable to high precision due to subtle satellite orbit 
perturbations, which are unknowable in real-time, and intentional 
signal instabilities  (dithering) imposed by the U.S. Department of 
Defense selective availability (SA)."

Could you substantiate this claim? Is there any reason a spoofer
could not do the following?

1. Set up a receiver near (within a 1000 miles or so) of the site he
is trying to spoof (he needs to be able to see roughly the same
satellites as the spoofed site).

2. Extract from the signal he receives the psudo-random sequence
sent from each satellite.

3. Buffer the sequence  from each satellite for a brief period (the extreme
case is 42 msec, the time it takes light to travel the diameter of the
earth).

4. Re-assemble the aggregate signal with the appropriate delays for the
location he is trying to spoof. These delays can be pre-computed.

5. Modulate a simulated 17 cm carrier appropriately to produce a synthesized 
signal, identical to that received at the location to be spoofed, and use
that to fake his location.

I note that there is commercially available test equipment to simulate
GPS satellites, which transmit a signal appropriate for any location you 
dial into them (btw, setting one of these up near a location using your
protocol  leads to an interesting denial-of-service attack, since you
can overwhelm the satellite signal with a false one giving a bogus
location).

Even if the spoofer has to extract sequence data from the real satellites,
the storage requirement is not  onerous, since he already knows just
how long he has to delay each satellite's signal, and need buffer only
the appropriate chunk for each satellite.

While this *does* result in a slight delay while the sequence data for
each satellite is gathered, the extreme case is 42 ms. Internet
transmission delays on the order of 100 ms are common, so your
system will have to accept location data which is this old.

The computational load neccesary to spoof the signal is not excessive -
it's essentially the reverse of that used to extract location data from the
aggregate  signal, a process which is not a one-way function.

Thus, the signal can be spoofed.

Second, your system broadcasts potentially sensitive location data.
Your protocol will be a gold mine for the traffic analysts  The
data can be used for later spoofing attacks, or, in an 
operational situation, to target munitions.

While encrypting the link can protect this data, if you can use encryption,
you can also use digital signatures  for authentication.

Finally, GPS receivers don't work too well in steel-framed 
buildings. There are substantial shielding and multipath problems
( for your system to work, the antenna needs to be near
the originating node, not on the roof). You do not appear to address
these problems. 

Peter Trei
trei@process.com

--------------------------------------------------------
I append the original article, as it appeared on the
list on April 10th.
--------------------------------------------------------


Location-based System Delivers User
Authentication Breakthrough

By Dorothy E. Denning and Peter F. MacDoran
Copyright(c), 1996 - Computer Security Institute - All Rights Reserved
Top - Help



Existing user authentication mechanisms are based on information the user
knows (e.g., password or PIN), possession of a device (e.g, access token or
crypto- card), or information derived from a personal characteristic
(biometrics). None of these methods are foolproof. Passwords and PINs are
often vulnerable to guessing, interception or brute force search. Devices
can be stolen. Biometrics can be vulnerable to interception and replay.

A new approach to authentication utilizes space geodetic methods to form a
time- dependent location signature that is virtually impossible to forge.
The signature is used to determine the location (latitude, longitude and
height) of a user attempting to access a system, and to reject access if the
site is not approved for that user. With location-based controls, a hacker
in Russia would be unableto log into a funds transfer system in the United
States while pretending to come from a bank in Argentina.

Location-based authentication can be used to control access to sensitive
systems, transactions or information. It would be a strong deterrent to many
potential intruders, who now hide behind the anonymity afforded by their
remote locations and fraudulent use of conventional authentication methods.
If the fraudulent actors were required to reveal their location in order to
gain access, their anonymity would be significantly eroded and their chances
of getting caught would increase.

Authentication through geodetic location has other benefits. It can be
continuous, thereby protecting against channel hijacking. It can be
transparent to the user. Unlike most other types of authentication
information, a user's location can serve as a common authenticator for all
systems the user accesses. These features make location-based authentication
a good technique to use in conjunction with single log-on. Another benefit
is there is no secret information to protect either at the host or user end.
If a user's authentication device is stolen, use of the device will not
compromise the system but only reveal the thief's location. A further
benefit of geodetic-derived location signatures is that they provide a
mechanism for implementing an electronic notary function. The notary could
attach a location signature to a document as proof that the document existed
at a particular location and instant in time.

The use of geographic location can supplement or complement other methods of
authentication, which are still useful when users at the same site have
separate accounts and privileges. Its added value is a high level of
assurance against intrusion from any unapproved location regardless of
whether the other methods have been compromised. In critical environments,
for example, military command and control, telephone switching, air traffic
control, and banking, this extra assurance could be extremely important in
order to avoid a potential catastrophe with reverberations far beyond the
individual system cracked.

How it works

International Series Research (Boulder, CO) has developed a technology for
achieving location-based authentication. Called CyberLocator, the technology
makes use of the microwave signals transmitted by the twenty-four satellite
constellation of the Global Positioning System (GPS). Because the signals
are everywhere unique and constantly changing with the orbital motion of the
satellites, they can be used to create a location signature that is unique
to a particular place and time. The signature, which is computed by a
special GPS sensor connected to a small antenna, is formed from bandwidth
compressed raw observations of all the GPS satellites in view. As currently
implemented, the location signature changes every five milliseconds.
However, there are options to create a new signature every few microseconds.

When attempting to gain access to a host server, the remote client is
challenged to supply its current location signature. The signature is then
configured into packets and transferred to the host. The host, which is also
equipped with a GPS sensor, processes the client signature and its own
simultaneously acquired satellite signals to verify the client's location to
within an acceptable threshold (a few meters to centimeters, if required).

For two-way authentication, the reverse process would be performed. In the
current implementation, location signatures are 20,000 bytes. For continuous
authentication, an additional 20 bytes per second are transferred. Re-
authorization can be performed every few seconds or longer. The location
signature is virtually impossible to forge at the required accuracy. This is
because the GPS observations at any given time are essentially unpredictable
to high precision due to subtle satellite orbit perturbations, which are
unknowable in real-time, and intentional signal instabilities  (dithering)
imposed by the U.S. Department of Defense selective availability (SA)
security policy. Further, because a signature is invalid after five
milliseconds, the attacker cannot spoof the location by replaying an
intercepted signature, particularly when it is bound to the message (e.g.,
through a checksum or digital signature). Continuous authentication provides
further protection against such attacks.

Conventional (code correlating and differential) GPS receivers are not
suitable for location authentication because they compute latitude,
longitude and height directly from the GPS signals. Thus, anyone can report
an arbitrary set of coordinates and there is no way of knowing if the
coordinates were actually calculated by a GPS receiver at that location. A
hacker could intercept the coordinates transmitted by a legitimate user and
then replay those coordinates in order to gain entry. Typical code
correlating receivers, available to civilian users, are also limited to 100
meter accuracy. The CyberLocator sensors achieve meter (or better) accuracy
by employing differential GPS techniques at the host, which has access to
its own GPS signals as well as those of the client. DGPS methods attenuate
the satellite orbit errors and cancel SA dithering effects.

Where it works

Location-based authentication is ideal for protecting fixed sites. If a
company operates separate facilities, it could be used to restrict access or
sensitive transactions to clients located at those sites. For example, a
small (7 cm x 7 cm) GPS antenna might be placed on the rooftop of each
facility and connected by cable to a location signature sensor within the
building. The sensor, which would be connected to the site's local area
network, would authenticate the location of all users attempting to enter
the protected network. Whenever a user ventured outside the network, the
sensor would supply the site's location signature. Alternatively, rather
than using a single sensor, each user could be given a separate device,
programmed to provide a unique signature for that user. Location-based
authentication could facilitate telecommuting by countering the
vulnerabilities associated with remote access over dial-in lines and
Internet connections. All that would be needed is a reasonably unobstructed
view of the sky at the employee's home or remote office. Related application
environments include home banking, remote medical diagnosis and remote
process control. Although it is desirable for an antenna to be positioned
with full view of the sky, this is not always necessary. If the location and
environment are known in advance, then the antenna can be placed on a window
with only a limited view of the sky. The environment would be taken into
account when the signals are processed at the host.

For remote authentication to succeed, the client and host must be within
2,000 to 3,000 kilometers of each other so that their GPS sensors pick up
signals from some of the same satellites. By utilizing a few regionally
deployed location signature sensors (LSS), this reach can be extended to a
global basis. For example, suppose that a bank in Munich needs to conduct a
transaction with a bank in New York and that a London-based LSS provides a
bridge into Europe. Upon receiving the location signatures from London and
Munich, the New York bank can verify the location of the Munich bank
relative to the London LSS and the London LSS relative to its own location
in New York.

The technology is also applicable to mobile computing. In many situations,
it would be possible to know the general vicinity where an employee is
expected to be present and to use that information as a basis for
authentication. Even if the location cannot be known in advance, the mere
fact that remote users make their locations available will substantially
enhance their authenticity. In his new book, The Road Ahead, Bill Gates
predicts that wallet PCs, networked to the information highway, will have
built-in GPS receivers as navigational assistants. With the CyberLocator
technology, these PC receivers can also perform authentication while being a
factor of ten less expensive than conventional code correlating receivers
(most of the processing is executed in the host rather than the remote
units), which only achieve 100 meter accuracy, and a factor of a hundred
less expensive than conventional DGPS receivers. Location-based
authentication is a powerful new tool that can provide a new dimension of
network security never before possible. The CyberLocator technology is
currently operational in a portable demonstration.


Dorothy E. Denning is professor of computer science at Georgetown
University (Washington, D.C.) and consultant to ISR. She can be reached at
202-687-5703 or denning@cs.georgetown.edu. Peter F. MacDoran is president
and CEO of International Series Research, Inc. (Boulder, CO). He can be
reached at 303-447- 0300 or pmacdorn@isrinc.com.

$0$AD

Peter Trei
Senior Software Engineer
Purveyor Development Team                                
Process Software Corporation
http://www.process.com
trei@process.com