Eternity - an alternative approach

-----BEGIN PGP SIGNED MESSAGE-----

In one of the messages on the eternity thread someone suggested using
radio signals. This is an interesting approach which certainly deserves
further attention but digital radio requires special hardware which I prefer 
to avoid. However, some of the characteristics of radio signals can be 
emulated on the internet.

Radio links are perfect for hiding the location of receivers. The
equivalent for this is of course services like usenet and IRC. Both are
already used for providing anonymity.

Radio is also quite good at hiding the location of the transmitter - you
tune your radio and the signals are just there, coming down your antenna.
It takes special equipment to locate where they originated.

It occured to me that the equivalent on the net would be to receive
packets with invalid source addresses. They are just there, coming dowm
the phone line to your modem. It takes significant resources and snooping
on a massive scale to locate where they are coming from.. All this is
assuming you can find some way to send a request with your address to the
server.

For eternity this means that the location of servers where documents are
stored can be kept secret if they transmit the documents using packets
with invalid source addresses. Receivers of eternity documents will not
be able to tell where they are coming from.

Since there is no bidirectional link between transmitter and receiver the
protocol must be implemented over UDP. To protect against packet loss the
data may be encoded with redundancy. The correction code can be similar
to that used on RAID disk arrays - adding redundant blocks consisting of
XOR of other blocks. Missing blocks can be reconstructed by XORing
together other blocks received.

To provide even better anonymity the transmitter may send the messages
through one or more onion routers which decrypt one layer of encryption.
These routers do not store information, do not know what information
passes through them and do not know where it's coming from. Any packets
they forward are sent without a valid source address. This should make
them less vulnerable to attacks of all kinds. Can anyone speculate about
their liability under US law?

Putting caches on the routers will help both performance and security.
Will it affect their liability in any way?

If encryption is implemented properly all packets will look identical:
blocks of random data of the same size and no source address. This will
make traffic analysis difficult. The distance to the transmitter can be
hidden by initializing the packet's time-to-live field to a value with
random variations.

That was the easy part. The hard part is to get the requests to the
document servers without knowing their addresses. This can be done
through the equivalent of remailer reply blocks or some kind of broadcast
medium.

The reply blocks would be implemented by the onion routers. Each hop
could generate several copies, both for redundancy and as decoys.
Packets used in reply blocks can look identical to packets with document
data, including the use of invalid source addresses.

In the broadcast version clients and servers subscribe to channels with
many participants to hide their identity in the crowd. The system could
use an existing service like IRC or a dedicated network, possibly based
on modified IRC server code.

Sympathizers with bandwidth to spare can subscribe too to provide better
cover for the actual document servers. Servers could listen using trusted
proxies or chains of semi-trusted proxies. The division into channels
allows growth of the system if the total bandwidth of requests becomes
too high.

Using the public IRC networks has the advantages of being less suspicious
and not requiring the deployment of new infrastructure. Requests could
even be hidden steganographically in IRC traffic. Unfortunately, IRC
server operators are touchy about abuse of their systems.

Requests should be small. This allows replicating them to a large number
of receivers without taking too much bandwidth. They can also be made to
be of fixed size and contain no controversial information. Requests
contain the document index as a compact one-way hash, the requester's IP
address and the server's ID code. It may also contain hashcash or ecash
payment. Requests should be encrypted to the server's public key. To keep
them small, elliptic curve encryption may be used.

A server may send random cover packets to hide the correlation between
a request and its response. It may also add random delay before
responding to the request. Delay will also help limit the bandwidth.
Users of eternity need to be more patient that the average web surfer and
the protocol should use generous timeouts. Since UDP has no flow control
the server should always assume the client is using a modem connection
and limit its transmission speed accordingly. Security has a price.

The client software may be implemented as a local web proxy (in Java?)
which identifies requests to eternityspace and converts them to this
protocol. Public gateways may be set up to allow anyone to read eternity
URLs. These public may also be used to protect the anonymity of the
reader like the Anonymizer.

Document servers and routers must be located on networks where there is
no spoofed packet filtering. Since the common use of spoofed packets is
in denial-of-service attacks more and more filters are being installed.
It may become harder in the future to find places to put the servers and
routers. I believe it should always be possible since it is not practical
to implement filtering in places where there is a lot of traffic from
different sources. In cases where the document server is located on a
filtered network it may have to trust the routers to hide its location.

This system is far from perfect: it has too many components and the weakest link is the transmission of requests to servers. I still believe it has some interesting sides which deserve further discussion. Your ideas are welcome.

- -----------------
Kay Ping 
nop 'til you drop

finger kping@nym.alias.net for key
DF 6D 91 18 A6 59 41 96 - 89 01 69 B7 9D0 4 AE 53

-----BEGIN PGP SIGNATURE-----
Version: 2.6.3ia
Charset: cp850

iQEPAwUBNL26txHPAso8Qp7tAQGLEAfPeRgIL3OqiG67CLBO9TCe/oV5lmw66Pz1
Wl17ajHeSCX6qjACSZ3La73drUjIftL0G/18PkLd48VGmsF6izCnXB4fh8MAB6Ve
QWjhRTvRSkKkwXK4t2tx6CUCdxOaJ9Phd6J02Z+MdjEGJ3jAdUdaHWo5zM5i6Ris
wkgATEhGMpw8tjlnvR4erwu51iSrt62huPWJXl1pjyPfQbl0iyQtcGdQ1spIWLJC
oaOI7QchHK3LjSuzN54MVCjRdz8fiI6JHAUnlqqsW29LBOZQkSnQedORCuIALwqe
qOhSxMKbciLdVde3BBtpILpz5y91ulAecxOOwcnc7m5Pjw==
=iLP5
-----END PGP SIGNATURE-----

Kay Ping wrote on 1998-01-16 22:02 UTC:
> Radio links are perfect for hiding the location of receivers.

Actually, this is only true for extremely carefully shielded military
receivers and not for normal radios. Every receiver contains a local
oscillator to bring the signal down to intermediate frequency (IF), which
is emitting EM waves itself. In addition, the IF signal is emitted
as well.

As Peter Wright reported in his autobiography, British counterintelligence
(MI5) used vans and planes already in the 1950s to detect spys while
they received radio communication messages from Moscow and to protocol,
which frequency bands the embassies were monitoring (operation RAFTER).
Efficient receiver detection is an active process: You send out short
bursts of a wideband jamming signal and try to find the downtransformed
intermediate frequency equivalent of your burst in the compromising
emanations of the receiver. This way, you get not only the location of
the receiver, but also the precise frequency to which it is tuned.

Locating radio receivers within a radius of many hundred meters this way
was already state of the art in the spook community over 40 years ago,
so you can safely assume that with digital signal processing, the
performance parameters of modern systems have been increased
significantly. Sending out spread-spectrum style pseudo-noise signals
in the active probing bursts could give you in modern receiver detectors
a considerable signal gain.

Markus

-- 
Markus G. Kuhn, Security Group, Computer Lab, Cambridge University, UK
email: mkuhn at acm.org,  home page: <http://www.cl.cam.ac.uk/~mgk25/>

> Kay Ping wrote on 1998-01-16 22:02 UTC:
> > Radio links are perfect for hiding the location of receivers.
> 
> Actually, this is only true for extremely carefully shielded military
> receivers and not for normal radios. Every receiver contains a local
> oscillator to bring the signal down to intermediate frequency (IF), which
> is emitting EM waves itself. In addition, the IF signal is emitted
> as well.
> 
> As Peter Wright reported in his autobiography, British counterintelligence
> (MI5) used vans and planes already in the 1950s to detect spys while
> they received radio communication messages from Moscow and to protocol,
> which frequency bands the embassies were monitoring (operation RAFTER).
> Efficient receiver detection is an active process: You send out short
> bursts of a wideband jamming signal and try to find the downtransformed
> intermediate frequency equivalent of your burst in the compromising
> emanations of the receiver. This way, you get not only the location of
> the receiver, but also the precise frequency to which it is tuned.
> 
> Locating radio receivers within a radius of many hundred meters this way
> was already state of the art in the spook community over 40 years ago,
> so you can safely assume that with digital signal processing, the
> performance parameters of modern systems have been increased
> significantly. Sending out spread-spectrum style pseudo-noise signals
> in the active probing bursts could give you in modern receiver detectors
> a considerable signal gain.
> 
> Markus

Hi,

I talked to some RF guys about the RAFTER attack about a year ago.
Their opinion was that since modern receivers have GaAs FET mixers,
they don't leak the LO or IF out the antenna like the old fashioned
inductor based mixers did.

This should be trivial to confirm with a spectrum analyzer.

Eric

There has been quite a bit of discussion on the list of late regarding
covert radio communication. I've been doing some investigation and have
identified a promising low probablity of intercept technology for
terrestrial, ground-to-air and earth-to-satellite communication. I'd like
to assemble a small team to build and test a proof of concept. Anyone with
interest and skillsplease contact me?

--Steve

On Mon, Jan 19, 1998 at 01:55:55PM -0800, Eric Blossom wrote:
> > Kay Ping wrote on 1998-01-16 22:02 UTC:
> > > Radio links are perfect for hiding the location of receivers.
> > 
> > Actually, this is only true for extremely carefully shielded military
> > receivers and not for normal radios. Every receiver contains a local
> > oscillator to bring the signal down to intermediate frequency (IF), which
> > is emitting EM waves itself. In addition, the IF signal is emitted
> > as well.
> > 
> > As Peter Wright reported in his autobiography, British counterintelligence
> > (MI5) used vans and planes already in the 1950s to detect spys while
> > they received radio communication messages from Moscow and to protocol,
> > which frequency bands the embassies were monitoring (operation RAFTER).
> > Efficient receiver detection is an active process: You send out short
> > bursts of a wideband jamming signal and try to find the downtransformed
> > intermediate frequency equivalent of your burst in the compromising
> > emanations of the receiver. This way, you get not only the location of
> > the receiver, but also the precise frequency to which it is tuned.
> > 
> > Locating radio receivers within a radius of many hundred meters this way
> > was already state of the art in the spook community over 40 years ago,
> > so you can safely assume that with digital signal processing, the
> > performance parameters of modern systems have been increased
> > significantly. Sending out spread-spectrum style pseudo-noise signals
> > in the active probing bursts could give you in modern receiver detectors
> > a considerable signal gain.
> > 
> > Markus
> 
> Hi,
> 
> I talked to some RF guys about the RAFTER attack about a year ago.
> Their opinion was that since modern receivers have GaAs FET mixers,
> they don't leak the LO or IF out the antenna like the old fashioned
> inductor based mixers did.
> 
> This should be trivial to confirm with a spectrum analyzer.
> 
> Eric

	This varies a great deal.  Generally cheap vhf/uhf scanners and
the like radiate quite a bit of energy and can be easily seen on a
spectrum analyzer at hundreds of feet (with no special effort).   A good
bit of energy escapes many scanners through the cheap and poorly
shielded plastic case and power cords rather than leaving via the
antenna, so even if a good broadband preamp is used between the antenna
and the input of the scanner - which should effectively eliminate LO
radiation from going out the antenna because the preamp is going to have
to have a great deal of loss for energy routed through it backwards
(output to input) or it would be unstable and oscillate -  the signals
radiated from the radio itself and the power cord will give it away.

	First local oscillator energy is the most easily seen radiation
from scanners - IF radiation is much lower in level in most modern 
gear because of the very short lead lengths and comparitively low signal
levels and good decoupling from the antenna - IF frequencies (save the
first IF on many scanners) are low enough so the component leads don't
act like a very good antenna because they are such a tiny fraction of
a wavelength.

	High grade military/spook class receivers are much better
shielded, some in fact to TEMPEST level specs,  and don't have the
problems that most  cheap scanners for hobbiests have.  If used with a
good preamp ahead of them they are very hard to detect if the shielding
is intact and undamaged (which may or may not be the case for a unit
that has been kicked around for years and carelessly repaired and
modified).

	But a fair number of modern HF (2-30 mhz)  communications
receivers and transcievers use no or very little RF amplification before
the first mixer in order to maximize dynamic range and third order
intercept and their 1st LO's (usually in the low VHF range) are not as
well isolated from the antenna.

	However, most modern receivers use synthesized local oscillators
phase locked to a local crystal oscillator and the LO is not as likely
to be detectably modulated by the audio the receiver is receiving as was
the case with the vacuum tube era communications receivers in the RAFTER
era that used free running LC tuned oscillators.   Power supply regulation
and decoupling is much better in modern gear, and this combined with
the use of phase locked synthesized oscillators means that while it may
be possible to detect radiation from the receiver LO, it is not as easy
to detect fm and am sidebands coming from the receiver audio which was
the basis of a lot of RAFTER work.

	But sensitive spectrum analyzers are available and not uncommon
these days, so anyone who is trying to operate an undetected receiver
for any serious purpose now has the tools to determine just how much his
gear is radiating.   Of course the detection gear has gotten better too,
but careful shielding up to and including faraday cages, use of good
preamps and circulators and use of spectrum analyzers to check for stray
radiation makes it less likely that someone will  easily find a
carefully hidden receiver than in the past.   But a plastic cased
consumer grade scanner from Radio Shack may be detectable a half mile
away... or more...



-- 
	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

-----BEGIN PGP SIGNED MESSAGE-----

On Sat, 17 Jan 1998, Markus Kuhn wrote:

> Kay Ping wrote on 1998-01-16 22:02 UTC:
> > Radio links are perfect for hiding the location of receivers.

[...]

> As Peter Wright reported in his autobiography, British counterintelligence
> (MI5) used vans and planes

IIRC the BBC uses such vans to detect unlicened recivers in Brition.
(However I have also heard that such vans are more effective at spreading
FUD).

- -- 
Please excuse my spelling as I suffer from agraphia see the url in my header. 
Never trust a country with more peaple then sheep.  ex-net.scum and proud
You Say To People "Throw Off Your Chains" And They Make New Chains For
Themselves? --Terry Pratchett.

-----BEGIN PGP SIGNATURE-----
Version: 2.6.3i
Charset: noconv

iQCVAwUBNMWHZKQK0ynCmdStAQFukQQAqRsMoLTPkunuFQWoyzvG7vLtRNEh0Xgf
L/V/J9/O6PeCKXZhURMdR9FHQvsZ7ETS677LsuAdapoS+swwtRxWhwHBXdJkDa3M
Oyz23S4Q3QR5WtvfRz7gj843yjpig7Hm/mwVUMTmUXC+lp6TXmKJAH/vFHyOkxPl
CgFsMHGFUN0=
=DeXr
-----END PGP SIGNATURE-----