Yanek - I do think your dongle idea is promising, but there are a couple of other problems I haven't seen you mention.

The concept is by no means complete. I will continue to add improvements as people suggest them.

One problem with this is that the terminal programs I use tend not to just blast the whole mail message out in literal form, especially since it is likely to be longer than one screenful. Instead, it pauses after each screen, prompting with something like "type space for next page",

It would be transparent the other way, so that you would press space or 'n' or 'return' or anything you want.

or such. Also, it may insert VT100 control sequences as it goes from page to page in order to clear the screen, etc.

It should not be too difficult to ignore any codes that are not part of the data stream. If the message is coded with something like uuencode or base64 then the uniform-length lines of data should be clearly disinguishable from any other text such as "Press N for next screen" or an escape sequence. Also, overlapping display (some pagers show you a couple of lines from the previous screenful) could be dealt with by detecting duplicate lines.

...

When you are done, you press Q, and the plaintext is immediately deleted from the memory.

Well, I'd think you'd want the option of saving the plaintext to a file on your PC, not always deleting it from memory.

By memory I meant the RAM in the dongle. If you wanted to save it, you could have just used to "logifile" function of your communication software.

What if you wanted to do what I'm doing here, which is to quote the message you are replying to?

I have not thought about that. Right away I see several possibilities: one, if the message you are replying to is the one you just decrypted, it could have been stored in the buffer, and you can delete the irrelevant lines, and insert your comments. The other would be that as you read a message, you could write and encrypt your reply. In that case it would also be simple to automatically use the public key that was used to verify the signature. I am sure there are other ways. There is essentially no limit on possibilities once you have a CPU. I will think about it some more. Thanks for bringing this up.

...

simple line editor, that works with any terminal or terminal

I think this is OK, although I'm not sure how happy I'd be with your line editor, not being accustomed to using BBS's.

You could probably download an editor of your choice into it. There could be several editors available. Maybe even a simple screen editor that recognizes several most common terminals like vt100 and ansi.

like I can always compose on my PC, so that's OK.

If your comm software supperted such a feature, the dongle could transmit a magic sequence which would make the pc to automatically go into an editor and then transmit the text. I would imagine eventually the most popular communication programs would have this feature. Some might already. The point is that if you happen to be using a dumb terminal or some machine or software WITHOUT this capability you could still receive and send secure messages.

...

Then you may want to scan a newsgroup like alt.w.a.s.t.e for any mail

...

message whose key-id matches one of the keys on your key-ring, decrypts and displays it, while ignoring all the other messages

You might have a buffering problem here. A decryption takes 30 seconds

They would not be decrypted "on-the-fly". Each Subject: line would contain the MD5 hash of the public key. It would take a fraction of a second to determine if that hash matches one of your keys for which the hash values would be precomputed and stored. Then the message would be captured. Later, when the scan is complete, it would be decrypted and displayed.

reasons; first, you don't know what flow control the host uses (if any), and second, it would expose the fact that you had found a message you wanted to decrypt).

That is true. It should not be a problem though, if there are not too many messages for you. I plan for the device to have 256k of ram, so it could hold a few messages. The ram could be expanded if necessary.

You can't really anticipate how big the buffer might need to be, especially if this method of anonymous communication became popular. As I said in a previous post, you could easily have thousands of messages to sort through, with perhaps dozens for you.

I envision that this method would only be used for short and important messages, mostly of the form: "Contact me by sending mail to xyzzy@anon-r-us.com, key: dfgSDFgds34DF" So they would be short, easy to buffer, quick to decrypt and hopefully not too many of them. If there WAS too many messages, more than you were able to buffer, then it would just ignore the rest, process the ones it did get, and then you would request a second scan during which the first ones (that have already been processed) would be ignored, and the rest would be received. The problem with this is that the host would know that if you scan twice you have a lot of messages. They dont know which ones though. Or maybe you just had a transmission error...

solution than just encouraging people to buy laptops and use them?

Most people that use e-mail already have a computer, They may not want to buy another one. They may use the computer or terminal on their desk for other things too, and would not want the inconvenience of having to use a separate screen/keyboard anytime they receive or want to send a private message. Laptop will cost more. Why buy a screen, keyboard, disk drive if you already have it. Hopefully my device will be cheap. It should be cheaper than any laptop. All I have is two serial ports, a cpu, some RAM and support cirquitry. I don't have a full-fledged keyboard, display, disk (or card) drive, and all the other stuff that goes in a laptop. Laptops are bigger. Or if they are small, they have lousy screens/keyboards. I hope to make my device smaller than even the so-called "palmtops". It would be easier to carry with you, in your pocket. It would be easier to hide if necessary. You can't make a laptop. You cant easily open it up, or if you can, it is complex enough that you can not completely understand its design. You can make a dongle from a kit, testing each part before you assemble it. You could make one from scratch. You can design one if you don't trust the schematics. Or you could show it to a competent electronics engineer and he could tell you if it does what it says it will. A laptop is not designed with security in mind. If you find that someone has just learned your secret code (that you used to encrypt your private key) and are going to seize your laptop in 15 seconds, you can't immediately destroy the key (you need to turn it on, boot or resume, exit from whatever program you were running or create a new window, CD to the directory where you store your private key, delete the file, run some program that makes sure the sectors are cleared.) And you still can't be sure that the key is not stored in some buffer or disk or cpu cache or a temporary file. Compare this with punching in a short "DESTRUCT" sequence on the keypad; if you have the device in your pocket, you can even do it without taking it out of your pocket, if you know the keypad well; And be absolutely sure that all information is completely gone.

Despite my negative comments here, I think there is room for plenty of different approaches to making crypto easier to use.

Absolutely. I am not saying that a hardware device designed specifically for cryptography is the ONLY way to go, what I do believe is that it is just a very good way to make easy access to cryptography widespread. A person that does not even know enough to be able to download, uncompress and install a software package could get one of these devices and make use of it. And there is no limit as to what a knowledgeable person could do with such a device.

If the dongle really could be made available at the price level you mentioned, there might be some interest ... and if there was one for $200 or less that would definately be interesting to me.

Right now I hope to make it available for about $100. I don't know if I can yet, but will find out soon. -- Yanek Martinson mthvax.cs.miami.edu!safe0!yanek uunet!medexam!yanek this address preferred -->> yanek@novavax.nova.edu <<-- this address preferred Phone (305) 765-6300 daytime FAX: (305) 765-6708 1321 N 65 Way/Hollywood (305) 963-1931 evenings (305) 981-9812 Florida, 33024-5819