...

I can't violate my NDA, but PGPFone will be a *major* revolution for communication security. I would not be surprised to see several hundred thousands of users. PGP will plale in comparison.

I sincerely doubt this. if someone could find a way of doing voice encryption through simple idiotproof hardware adapters (I am thinking of cups that you could attach to any standard phone) the voice encryption will not be widely used, I think. there are very many PGP users right now, say at least in the tens of thousands if not hundreds of thousands. the requirements for live voice encryption are pretty significant: a fast computer and fast modem. this alone is only a subset of those people using PGP right now. also, I doubt there are going to be few people who use PGP phone but not PGP software.

I don't think CPUs and modems are an issue. When I bought my 486/DX2 more than a year ago, it was a near top of the line machine (a P66 was the only thing better and it was a lot more expensive). Now, I can look in computer shopper and see that not only is my machine not near the top of the line, it's not even "Entry Level". Pentium 75/90 systems are going for 1/3 the price I bought my computer for, and those systems have PCI buses (vs my VESA local bus), larger HDs, EDO RAM, faster video cards, etc. Assuming a baseline of a 486DX/33 or faster (like a DX4) is not unreasonable. Secondly, 14.4K modems are a dime a dozen. You can get them as low as $50 (with RPI) or $70-80 for full functionality. 28.8K modems can be bought for $150. The problem with PGPFone as I see it, is that it's an application and not a application to a protocol. To get voice encryption in large scale use will require several things IMHO 1) performance is reasonable 2) user interface is very easy to use, as easy as using a walkie talkie with a key 3) software is very easy to setup up (no knowledge of hayes commands required, no editing of slip configuration, etc) Finally, even that is not going to drive the system into a defacto ubiquitous standard unless 4) a complete, easy to read specification of the protocol used is published (perhaps as an RFC) 5) third party applications that use the protocol evolve. #5 is needed because competition between applications writers will improve the human interface of the software beyond what the PGP authors can design. (who are more likely algorithm specialists, not human interface people) 6) network independent this will be a benefit to people who want to make long distance calls over data networks. it could also be used by companies for secure teleconferencing. I would like to see a secure voice communication protocol that is divorced from the particular details of the algorithms used (although a base level of some voice compression technique + DES + RSA will have to be used) That way, new and better algorithms can be dropped in depending on the network used (modem, ipx, tcp/udp, etc) and the bandwidth required (CELP vocoder, MPEG-audio, lossless encoding, progressive PCM, etc) -Ray

On Fri, 11 Aug 1995, Vladimir Z. Nuri wrote:

when someone invents cheap hardware that you can just plug on top of any existing phone, i.e. "the phone adaptor", TPA?, *that's* when the world is going to go crazy with crypto.

It's closer than you think. I've been messing with TI's Digital Signal Processing DSK. For $99 you get a DSP with audio in, audio out and 10k of memory. Reference implementations of : DTMF encoders/decoders; 300, 1200, 2400 baud modem programs; and voice processing software already exist. The TI Linear Products Transmission, Switching, Subscriber, and Transient Suppressors Data Book is sitting on my desk along with the Data Transmission and Control Circuits (etc) Data Book. Combined with the pinouts and software that came with DSP DSK I've been painfully trying to piece together how one might glue up a telephone interface. Unfortunately, I'm a software guy and am still learning. In any case, assuming Joe STUD Hardware Guy was willing to make a daughterboard with a 64k memory module (10k is just not enough) with an FCC approved telephone interface (available from parts suppliers). You could have a 25 MIPS based personal STU. I'm envisioning a 'black box' with two RJ-11 jacks which daisy chained in with your phone like a modern modem. Triggering key exchange could be a simple DTMF sequence. Ultimately, the device could be combined into a single board eliminating parts, reducing space, and lowering power supply requirments. The DSP DSK currently needs ~14 Volts AC. I'd like to see that black box portable and able to use some nine volt or AA batteries.

all this stuff that requires lots of hardware can be used, but the thing that will cause *everyone* to use it is when someone invents a TPA. I'm very, very surprised that no one has tried to do this yet.

In any case, I'm trying to do it, but my efforts will never realise a commericial result. I just don't have the money to buy parts in bulk, and without doing that units would be $200-$500 each.