Tunny-era cryptographers, faced with day-by-day cryptanalysis decisions, viewed 1940s-era crypto as “clean” – speaking 50 years later. They have had time to reflect – how close a call it was. It was only the flaws in the operations of all the modern machine models that made them susceptible to machine attack. Inducing such flaws (and building them in to the mental model of the typical operator, designer, and analyst) became part of the game. The weakest link is the human being; and to him (and its always him) you play a line. When he steps out of line, you crack the whip.
If I was an American designer, I’d spend money getting everyone to concentrate on their PC. Don’t worry about the cable box on a high bandwidth cable connection to the Central Office, sitting there in the corner listening to every blip produced by the PC’s electronics.
Using one (targeted) commodity device to *amplify* other device signals in its locality – to suit the signals collector array – is normal. It’s known as a sensor network. The induction loops at the typical US road intersection are another example. It’s not hard to count how many folks are in the car, and which devices they have in their possession.
But, even within the (well-duped) analytical mental models of the Callas’s of the world, its hard to satisfy the (somewhat artificial) quality metrics. The first, pre-production Fortezza cards (known as Tessera cards before a name change) had flaws in the random number handling of the device password – the root of the physical-crypto tree. But, then what do you expect for a $100 device?