[dayglored]

> One-time pad ... the key-exchange problem... It requires a true random number source... but this is easy using something like diode noise.

There's an easier way, assuming one initialization.

Regularly exchanging secret keys is risky; that's the whole point of public-key encryption. One-time pads should never have to be exchanged, but rather they should be independently derived at each end.

A good one-time pad (like your suggestion of diode noise) can be derived from a commercial live broadcast which is available to both the sending and receiving parties. For example (this is a very weak way to use it, for explanation purposes only):

• Compile a list of the songs played on a certain music radio station between (say) 9PM and 10PM on a certain evening. This merely requires listening to the radio for an hour and noting the song titles.
• For each song, take the first two letters of each word in the first two lines of the lyrics.
• Concatenate all those letters. An hour's worth is about 15 songs; each line is perhaps 8 words, so that's 32 letters times 15 songs = 480 characters, or about 2880 bits of key.
• Need a longer key, take more letters, or more lines, etc.

A better way to do it is to digitize the audio of a live broadcast (suitably low-pass filtered first, of course) and use that digital stream as the key.

The idea is to derive the key from something agreed-upon in one initialization, and which never has to be exchanged again. Subsequent broadcast times, stations, etc. can be encoded into messages, or perhaps posted in an innocuous third-party forum.

[Bobalu]

Good points.

What is nice is that the low price of thumb drives means that a random number sequence of many billions of bytes is easily placed on two thumb drives for near zero cost.

They would provide a lifetime of absolutely protected secure text exchanges.

And so long as the used bytes are securely wiped there is ZERO chance of the messages being decrypted.