From: jkreznar@ininx.com (John E. Kreznar)
Date: Sun, 10 Oct 93 23:19:47 PDT
To: cypherpunks@toad.com
Subject: Re: Spread-spectrum net (vulnerability of)
In-Reply-To: <9310090132.AA08927@dink.foretune.co.jp>
Message-ID: <9310110617.AA05506@ininx>
MIME-Version: 1.0
Content-Type: text/plain



> You can put up such a network, but given that your boxes have to receive
> the signal, it will be detectable (but not understandable) to 3rd parties.

Not necessarily.  Done right, direct sequence spread spectrum transmissions
cannot feasibly be detected except by a receiver using the right code.  To be
sure, ``feasibility'' here is being used in a weaker sense that cryptologists
are accustomed to, but the effect is similar.  Here are selected passages from
"Modern Communications and Spread Spectrum" by George R. Cooper and Clare D.
McGillem (McGraw-Hill, 1986) pp 309--311:

	...another objective in using spread-spectrum techniques is to make it
	more difficult for an unauthorized observer to detect the presence of
	the signal....

	When an unknown spread-spectrum signal is to be detected, it is
	necessary to assume something about what is known.  In particular it is
	assumed that the spread-spectrum code is not known, and that it is
	desired only to determine the existence of the spread-spectrum signal
	and not to decode it.  This implies then that some form of noncoherent
	detection must be employed and that the existence of a spread-spectrum
	signal can be determined only by noting a change in energy between the
	signal present and the signal not present.  Thus it is necessary to
	consider an energy detector. ...

[Engineering mathematics omitted.]

	...However, in the case of noncoherent detection, the detectability
	depends upon the square of [a specific signal-to-noise ratio that they
	define], rather than upon the first power; thus operating with small
	values of E/N_0 makes things more difficult for the interceptor than it
	does for the desired signal receiver.

> Anyone who wants to take you down will only need (1) a detector that can
> point out your boxes and (2) a small caliber rifle.

First he may need (0) the spread-spectrum code.  These can be made strong, in a
sense similar cryptographically strong random number generators.

> Since the cost to find and destroy is much less than the cost to make and
> deploy, a covert network of this sort wouldn't last long.

This cost comparison might be reversed by forcing the adversary to use
incoherent detection.

	John E. Kreznar		| Relations among people to be by
	jkreznar@ininx.com	| mutual consent, or not at all.