Re: Streams, Voice, and Sensitive Dependence

[Michael Motyka <mmotyka@lsil.com>]

> It is my understanding that the PGP algorithm is primarily intended
> as a block cypher.  Therefore, wouldn't an Outer Block stream cypher
> be more effective for phone conversations?  Both solutions require 
> hardware for a practical (tolerably noiseless) implementation, so 
> there's nothing to be gained by streaming data into a block format.
>
Unless you have a really fast ( 1 Mbit / s? ) data connection you're
going to want to do some voice compression. The algorithms I've seen
break the audio into discrete time frames and (de)compress frame by
frame. As a point of reference say about 16 bytes for every 33 msec of
voice. Quality roughly follows data rate, of course. This makes a block
cipher seem not so unreasonable.

Block cipher or stream cipher, either way you're going to have to
introduce a latency of _at_least_ a couple of frames to allow for
resends or deliberate out-of-order frame transmission. This makes the
block cipher look like the better choice.

I think that using HW voice compression and a 33.6 modem you could get a
full duplex secure conversation over POTS with a latency in the 0.1 -
0.3 second range and a direct cost in the vicinity of $100. With a
reasonably quick microP any encryption method could probably be done as
SW.

This is not a particularly difficult device to build. Any fine US
citizens want to build some prototypes? 

Mike