1.You've got 16 hardware devices that each generate random noise. 2.One of the devices fails (or is sabotaged) and emits a predictable stream (10101...) 3.The other 15 devices are just fine, and the stream generated by one device does not effect the stream of another. 4.You do not know of the (failure/sabotage) until *after* you've generated your encyrted documents and they are out of your hands.
So the revamped question is: How secure are those documents now?
Bottom line: You have lost some possible decryptions of the cyphertext. Every 16th bit is now determined therefore each 2nd byte has only 128 possible states. This, rather suprisingly, does not seem to affect the security (though I would say this is a flawed conclusion as we have a pile of sand problem here, how many grains make a pile, how many rngs can be flawed before the security is affected?) I really haven`t had the time to look at it properly, I will do so in the near future as it seems to be an interesting problem... Datacomms Technologies web authoring and data security Paul Bradley, Paul@fatmans.demon.co.uk Paul@crypto.uk.eu.org, Paul@cryptography.uk.eu.org Http://www.cryptography.home.ml.org/ Email for PGP public key, ID: 5BBFAEB1 "Don`t forget to mount a scratch monkey"