On 09/11/2013 07:18 PM, Perry E. Metzger wrote:

The attraction of methods that use nothing but a handful of
transistors is that they can be fabricated on chip and thus have
nearly zero marginal cost. The huge disadvantage is that if your
opponent can convince chip manufacturers to introduce small changes
into their design, you're in trouble.

And this is the reason that I'd be in favour of "diversity" -- using sound cards, lava-lamps, etc, etc. Sources that don't explicitly identify themselves
  as "the random number generator".

There's no way for a bad actor to cover "all the bases", and since these things are primarily used for things other than random-number sources, it may be hard to "break" them in ways that doesn't also break their primary purpose (although, if you're just mucking with the low-order "noise bits" of some arbitrarily-chosen digitization of a real-world source, it would be hard to tell the difference).

Marcus Leech
Principal Investigator
Shirleys Bay Radio Astronomy Consortium

The cryptography mailing list

Reply via email to