You can hide from the Chip, but not from the Man.

After reviewing the NIST rebuttal to Matt Blaze's Paper, 'Protocol Failure
in the Escrowed Encryption Standard', referring to how the Unit ID (UID)
was expanded from 24 bits to 32 bits, I e-mailed the following question to
Dorthy Denning, informing her that I wanted to share the answer.

The question arises, does the unit ID indeed contain a field registered to
the equipment manufacturer?

Professor Denning replied:

"Yes, the UID contains bits that identify the manufacturer."

(I didn't think to ask how many)

The implication is that a counterfeit LEAF is detectible.  As per FIPS Pub
185, The Escrowed Encryption Standard, a transmission or stream of data is 
preceded by the Cryptographic Protocol Field (CPF) which is registered to a 
particular application (Clipper phone - AT&T, for example).  The CPF is used
to determine where to find the LEAF, the LEAF Creation Method (LCM) and the
Family Key (KF).  Thus the CPF also identifies the manufacturer, or group of 
manufacturers for a theoretically second sourced product, by identifying the
data protocols of the encrypted data (RCELP in the case of AT&T). 

A Bogus LEAF tested against the Escrow Authenticator (EA)  must still match
the manufacturer information found in the Unit ID.  I would expect that
there is between 10 and 12 bits of the UID specifying manufacturer.

The bad news is that to escape detection by the Law Enforcement/National
Security monitoring activity, you need to produce a LEAF that not only
produces an acceptable Escrow Authenticator used by the recipient EES
chip, but also produces a UID falling with some number of bits that matches
LE expectations as a result of examing the CPF.

The problem is that without knowledge of the Family Key and the LEAF
creation method, there is no possiblity of checking for a match in the
UID's manufacturers identifier.

koontzd@lrcs.loral.com (David Koontz ) writes:

>The implication is that a counterfeit LEAF is detectible.  As per FIPS Pub
>185, The Escrowed Encryption Standard, a transmission or stream of data is 
>preceded by the Cryptographic Protocol Field (CPF) which is registered to a 
>particular application (Clipper phone - AT&T, for example).  The CPF is used
>to determine where to find the LEAF, the LEAF Creation Method (LCM) and the
>Family Key (KF).  Thus the CPF also identifies the manufacturer, or group of 
-------------------^^^^
>manufacturers for a theoretically second sourced product, by identifying the
>data protocols of the encrypted data (RCELP in the case of AT&T). 

I am confused by the word "thus".  None of the three things in the CPF
mentioned in the previous sentence (where to find the LEAF, the LCM,
the KF (BTW, I thought the family key was a big secret?)) include the
manufacturer or the data protocols in any apparent way.  Are there more things
in the CPF than the three you listed?

Also, isn't it likely that RCELP will be widely used by all manufacturers
to be compatible with AT&T, so in practice all will use the same protocol,
and so this does not really identify the manufacturer?

As for recognizing bogus LEAF's, this would be only after decrypting with
the family key, right?  This is not supposed to be done routinely, although
it doesn't require access to the escrow database.  It's true that if a
family-key-decrypted LEAF using Blaze's rogue technique "stands out", that
certainly could call unwelcome attention to the users of his ideas.

Hal