Re: Faster way to deescrow Clipper

[koontzd@lrcs.loral.com (David Koontz )]

>From: sidney@taurus.apple.com (Sidney Markowitz)
>
>Could someone please enlighten me on this: It seems from the descriptions
>of the hack to fake a LEAF that 1) When two Clipper chips are going to
>communicate, one of them generates the session key and sends a LEAF to the
>other chip, 2) The second chip recognizes the LEAF as being valid based on
>the validity of the checksum, but does not determine the session key from
>the LEAF.

The session key encoded with the unit key in the Law Enforcement Access Field
is not recoverable.  The unit key is only contained within the orignating
clipper chip, and within the two escrowed key portions.

>If that's the case, then 1) How does the second chip find out what the
>session key is?

An external key exchange in performed.  In the case of AT&T TSD 3600s the
key is negotiated between them blindly.

>                 2) Doesn't the second chip also have to generate and send a
>LEAF, if for no other reason than to identify itself to the wiretappers,
>and if so won't that give away the session key if that chip's device is not
>also hacked?

FIPS Pub 185 requires that clipper phones adhere to a protocol not contained
within the FIPS, and requires transmission of the LEAF.

>              3) If all that is needed for this hack is a LEAF with a proper
>checksum, why go through the brute force method of generating random LEAFs?
>Why not just buy (or steal or whatever) another Clippered device that you
>never use for real communication so the wiretappers have no record of who
>has that serial number, and get LEAFs from it? For that matter, why can't
>you obtain one LEAF from listening to anybody's Clippered transmission and
>use it over and over again?

Using a constant unit ID even if the session key is not recoverable, still
leaves you open to traffic flow analysis.  Using a LEAF from another
clipper chip still identifies you (or serves to flag interest).  It would
be a wise to assume that the adversary routinely decodes all LEAFs crossing
their purview, recovering unit IDs.  (Which would at least flag a lot of
counterfeited LEAFS, were ID ranges or (as in IP) prefixes used.)