Re: Karn's note...

[karn@qualcomm.com (Phil Karn)]

I tend to doubt that the spooks have voice recognition technology in
regular widespread use, at least not the kind of ultra sophisticated
stuff that AI types seem to dream about.  It's possible that they use
less sophisticated stuff as a "pre filter" (compress out silence,
perhaps distinguish male from female voices, etc), but I'm sure that
the bulk of the work is still very labor intensive. Tens of thousands
of clerks, intercept operators and natural language translators have
long been employed by the NSA and there don't seem to be mass layoffs
of these sorts of people around Fort Meade.

And sophisticated voice recognition really isn't necessary when you
consider all of the information that cell phones and base stations
emit that is almost trivially processed automatically by an intercept
device: electronic serial numbers, Mobile Identification Numbers
(telephone numbers), handoff messages, channel assignment messages,
etc. It's no big deal at all to build boxes that automatically
intercept all calls made to or from a specific phone, assuming you
have an RF path to the target (e.g., from a car tailing a suspect).

As a manufacturer of cellular telephones, we have such a box
(commercially made by IFR) in our lab. We use it to test our phones in
their FM/analog mode.  The spooks (NSA and otherwise) simply cannot be
uninterested in boxes like these -- and in preserving their
capabilities.

One point I keep making about Clipper: it makes this sort of automated
identity tracking as easy on regular telephone lines as it already is
on cellular, because the chip serial number in the Law Enforcement
Block can be decrypted with just the (common) Family Key - you don't
need the escrowed keys. And sometimes simple traffic analysis can be
almost as deadly as getting the actual contents of a conversation.

Phil